EPA-600/3-76-037
April 1976
Ecological Research Series
TROPHIC CLASSIFICATION OF LAKES
USING LANDSAT-1 (ERTS-1)
MULTISPECTRAL SCANNER DATA
Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Corvallis, Oregon 97330
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ECOLOGICAL RESEARCH series. This series
describes research on the effects of pollution on humans, plant and animal
species, and materials. Problems are assessed for their long- and short-term
influences. Investigations include formation, transport, and pathway studies to
determine the fate of pollutants and their effects. This work provides the technical
basis for setting standards to minimize undesirable changes in living organisms
in the aquatic, terrestrial, and atmospheric environments.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/3-76-037
April 1976
TROPHIC CLASSIFICATION OF LAKES
USING LANDSAT-1 (ERTS-1)
MULTISPECTRAL SCANNER DATA
By
D. H. P. Boland
Assessment and Criteria Development Division
Con/all is Environmental Research Laboratory
Corvallis, Oregon 97330
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
CORVALLIS ENVIRONMENTAL RESEARCH LABORATORY
CORVALLIS, OREGON 97330
-------
DISCLAIMER
This report has been reviewed by the Corvallis Environmental Research
Laboratory, U.S. Environmental Protection Agency, and approved for
publication. Mention of trade names or commerical products does not
constitute endorsement or recommendation for use.
n
-------
PREFACE
In this report the satellite under scrutiny ts referred to as LANDSAT-1,
an acronym for Land Satellite One. It was the first space observatory
launched under the National Aeronautics and Space Administration's Earth
Resources Technology Satellite Program and originally carried the desig-
nation Earth Resources Technology Satellite One (ERTS-1). The series
name change was accomplished with the launch of ERTS-2 (j_.e.., LANDSAT-2)
in January 1975.
Both LANDSAT-1 and LANDSAT-2 are still operational at this point in time
(November 1975). The third satellite of the series, LANDSAT-3, is tenta-
tively scheduled for launch in February 1977. While this report was of
necessity limited to LANDSAT-1, the successful launch of LANDSAT-2, the
decision to proceed with LANDSAT-3, and the forthcoming Space Shuttle
Program will provide many opportunities to further define the role of
orbital-level remote sensors in surveying the earth's lacustrine resource
-------
ABSTRACT
This study evaluates the Earth Resources Technology Satellite One (ERTS-1;
I.e.., LANDSAT-1) multispectral scanner (MSS) as a means of estimating la-
custrine trophic state and the magnitude of two trophic state indicators.
Numerical classificatory methods are employed to ascertain the trophic
character of 100 lakes in Minnesota, Wisconsin, Michigan, and New York us-
ing the trophic indicators: chlorophyll a_, total organic nitrogen, inverse
of Secchi disc transparency, conductivity, total phosphorus, and an algal
assay yield. A complete linkage clustering algorithm is first used to ex-
amine the lakes for natural clusters. The resultant clusters are not read-
ily interpretable in terms of the three classic states of trophy (oligo-
trophic, mesotrophic, eutrophic). The hyper-dimensional cloud of data
points is then reduced in dimensionality through the ordination technique
of principal components analysis. A multivariate trophic state index (PCI)
is derived from the principal component analysis.
A binary masking technique is used to extract lake-related MSS data from
computer-compatible tapes. Data products are in the form of descriptive
statistics and photographic concatenations of lake images.
MSS color ratio regression models are developed which are of practical
value for estimating lake Secchi disc transparency and chlorophyll a_
levels for one date of LANDSAT-1 coverage. The trophic state of lakes,
as defined by lake position on the first principal component axis (PCI
value), is predicted using MSS color ratio regression models. Each date
of LANDSAT-1 coverage has its unique regression models for the prediction
of the trophic indicators and the trophic index; the models for different
dates vary greatly in their practical applicability.
The mean infrared-one band intensity levels for several "hypereutrophic"
lakes exceed their mean red band intensity levels and effectively isolate
them from other lakes in three-dimensional MSS color models.
A Bayes Maximum Likelihood Multispectral Classifier is employed to classify
a group of Wisconsin lakes using MSS colors in conjunction with the lakes'
trophic state index values. The results are depicted in the form of both
gray-scale and color-coded photographic concatenations.
The utility of the LANDSAT-1 MSS is most apparent when the seasonal con-
trasts between lakes at different points on the trophic scale are at a
maximum. Periods of excessive cloud cover, frames with faulty or missing
MSS data, and the need for some ground truth, impair, but do not preclude
its use in lake monitoring and classification. The use of computer-
compatible tapes in conjunction with digital image processing techniques
is essential if the maximum benefits are to be derived from the LANDSAT-1
MSS in lake-orientated studies.
iv
-------
CONTENTS
Preface i j -j
Abstract iv
List of Figures viii
List of Tables xii
Acknowledgements xviii
Sections
I CONCLUSIONS 1
II RECOMMENDATIONS 2
III INTRODUCTION 4
Statement of Purpose 4
Format of Report 5
The Earth Resources Technology Satellite 5
Orbit Parameters and Earth Coverage 6
Instrumentation 8
Products 12
Lake Monitoring Potential 12
Description of the Study Area 13
Geographic Area 18
Climate 18
Geology, Soils, and Land Use 18
Lake Selection Criteria and Location 18
Ground Truth Collection 23
Lake Site Selection 23
Lake Sampling Methods 23
Analytical Methods 28
IV LAKE CLASSIFICATION 30
Lakes as Natural Resources 30
Lake Succession and Eutrophication 31
Lake Succession 31
The Concept of Eutrophication 33
Trophic State and Trophic Indicators 35
Multivariate Classification of Lakes 37
Cluster Analysis 38
Objects and Attributes 38
Cluster Method 39
Results and Discussion 42
Principal Components Ordination 44
Methodology 47
Results and Discussion 48
Summary 55
-------
Sections Page
V LANDSAT-1 MSS DATA EXTRACTION TECHNIQUES AND PRODUCTS 57
MSS Data Extraction Approaches 57
Photographic Approach 57
CCT Approach 58
CCT Data Extraction Techniques 59
Image Processing System 59
Data Extraction Techniques 60
Digital Image Enhancement Techniques 68
Color Ratio Technique 68
Linear Contrast Stretching Technique 68
VI LANDSAT-1 MSS TROPHIC INDICATOR RELATIONSHIPS 72
Optical Properties of Pure and Natural Waters 72
Peripheral Effects 76
Relevant Remote Sensing Literature 82
Relevant LANDSAT-1 Investigations 84
Trophic Indicator Estimation 85
Lake Area Estimation 90
Secchi Disc Transparency Estimation 92
Chlorophyll a_ Estimation 93
VII LAKE CLASSIFICATION USING LANDSAT-1 MSS DATA 98
Trophic State Index Prediction Using MSS Data 98
PCI-MSS Regression Analyses 99
Three-Dimensional MSS Color Ratio Models 108
Lake Classification Using MSS Data in Conjunction with
Automatic Data Processing Technique 115
VII GENERAL SUMMARY 123
Lake Classification Using Ground Truth 124
MSS Estimation of Lake Area and Selected Trophic State
Indicators 125
MSS Prediction of Lacustrine Trophic State 126
Extraction Techniques and Products 126
LANDSAT-1 Coverage and Quality 127
IX REFERENCES 128
X LIST OF ACRONYMS AND SYMBOLS 147
XI APPENDICES 150
APPENDIX A Trophic Indicator Data for 100 NES-Sampled
Lakes 151
APPENDIX B Sampling Dates for 100 NES-Sampled Lakes 154
APPENDIX C Morphometry and Hydrology of Study Lakes 156
APPENDIX D LANDSAT-1 MSS Models, Concatenations, Area!
Relationships, and Descriptive Statistics 161
vi
-------
Sections
APPENDIX Dl
D2
D3
D4
D5
D6
D7
D8
D9
D10
Dll
D12
APPENDIX E
APPENDIX F
9 August 1972 (1017-16091, 1017-16093) 162
28 August 1972 (1036-16152) 166
11 June 1973 (1323-16100, 1323-16094) 170
17 July 1973 (1359-16091, 1359-16094) 175
14 August 1972 (1022-16373) 181
6 October 1972 (1075-16321) 189
8 October 1972 (1077-16431) 197
28 May 1973 (1309-16325) 204
3 July 1973 (1345-16322) 210
4 July 1973 (1346-16381) 216
19 August 1972 (1027-15233) 222
11 October 1972 (1080-15180) 228
N x N Squared Euclidian Distance Matrix 233
Listing of McKeon Cluster Analysis Program 240
-------
LIST OF FIGURES
Figure Pa9e
1 LANDSAT-1 ground coverage pattern 7
2 Schematic diagram of the LANDSAT-1 MSS scanning
arrangement 10
3 Ground scan pattern for a single MSS detector 11
4 Reproduction of an EROS Data Center IR2 print of Frame
1017-16093 (9 August 1972) 14
5 Reproduction of EROS IR1 print of Frame 1017-16093
(9 August 1972) 15
6 Reproduction of EROS RED (MSS Band 5) print of Frame
1017-16093 (9 August 1972) 16
7 Reproduction of EROS GRN (MSS Band 4) print of Frame
1017-16093 (9 August 1972) 17
8 Physical subdivisions 19
9 Geology 20
10 Distribution of principal kinds of soils: Orders,
Suborders, and Great Groups 21
11 Major land uses 22
12 Location of the study lakes 27
13 Hypothetical productivity growth-curve of a hydrosere 34
14 N x N S-matrix 41
15 Dendogram of 100 lakes sampled by the National
Eutrophication Survey during 1972 43
16 Geometrical interpretation of the principal components
for a hypothetical bivariate system 46
17 Three-dimensional principal component ordination of
100 lakes sampled by the National Eutrophication Survey
during 1972 56
VI 1 1
-------
Figure Page
18 Left half of MSS Frame 1017-16093 reproduced using IR2
DN values 61
19 An extracted section of LANDSAT-1 MSS Frame 1017-16093 62
20 First-stage MSS Band 4 (GRN) cleanup picture of Lake
Koshkonong 64
21 First-stage MSS Band 5 (RED) cleanup picture of Lake
Koshkonong 65
22 Four band (GRN, RED, IR1, IR2) concatenation of Lake
Koshkonong 66
23 IR2 concatenation of 15 lakes extracted from LANDSAT-1
MSS Frame 1017-16093 67
24 Lake Koshkonong image enhancement using six LANDSAT-1
MSS ratios 69
25 Contrast stretched images of Lake Koshkonong
(9 August 1972) 70
26 Contrast stretched images of Lake Koshkonong
(28 August 1972) 71
27 Reflectance curve for distilled water 74
28 Reflection characteristics of filtered and unfiltered
water samples from two Wisconsin lakes in the area of
Madison 75
29 Spectral distribution of solar energy 78
30 Solar zenith and solar elevation relationship 79
31 Percentage reflectance of the air-water interface as a
function of the angle of incidence measured from
normal direction 80
32 Comparison of the reflectance of chlorophyll-containing
plants with the attenuation length of sunlight in
distilled water 95
33 Three-dimensional color ratio model for 9 August 1972 110
34 Three-dimensional color ratio model for 11 June 1973 112
ix
-------
Figure Page
35 Three-dimensional color ratio model for 17 July 1973 113
36 Three-dimensional color ratio model for three dates
of LANDSAT-1 MSS coverage (9 August 1972, 11 June 1973,
17 July 1973) 114
37 ADP-classified lakes (9 August 1972), using 19 gray-
levels, one for each class 120
38 A color-enhanced version of the 19-class classification
of the 20 lakes (9 August 1972) 121
Dl-1 IR2 concatenation of 5 Wisconsin lakes extracted from
Frame 1017-16091 (9 August 1972) 163
D2-1 IR2 concatenation of 6 Wisconsin lakes extracted from
Frame 1036-16152 (28 August 1972) 167
D3-1 IR2 concatenation of 12 Wisconsin lakes extracted from
Frame 1323-16100 (11 June 1973) 171
D3-2 IR2 concatenation of 11 Wisconsin lakes extracted from
Frame 1323-16100 (11 June 1973) 172
D4-1 IR2 concatenation of 15 Wisconsin lakes extracted from
Frame 1359-16094 (17 July 1973) 176
D4-2 IR2 concatenation of 4 Wisconsin lakes extracted from
Frame 1359-16091 (17 July 1973) 177
D5-1 Three-dimensional MSS color ratio model of 12
Minnesota lakes extracted from Frame 1022-16373
(14 August 1972) 185
D5-2 IR2 concatenation of 12 Minnesota lakes extracted from
Frame 1022-16373 (14 August 1972) 186
D6-1 Three-dimensional MSS color ratio model of 15
Minnesota lakes extracted from Frame 1075-16321
(6 October 1972) 192
D6-2 IR2 concatenation of 8 Minnesota lakes extracted from
Frame 1075-16321 (6 October 1972) 193
D6-3 IR2 concatenation of 8 Minnesota lakes extracted from
Frame 1075-16321 (6 October 1972) 194
-------
Figure page
D7-1 Three-dimensional MSS color ratio model of 10
Minnesota lakes extracted from Frame 1077-16431
(8 October 1972) 200
D7-2 IR2 concatenation of 10 Minnesota lakes extracted
from Frame 1077-16431 (8 October 1972) 201
D8-1 Three-dimensional MSS color ratio model of 13
Minnesota lakes extracted from Frame 1309-16325
(28 May 1973) 207
D8-2 IR2 concatenation of 13 Minnesota lakes extracted
from Frame 1309-16325 (28 May 1973) 208
D9-1 Three-dimensional MSS color ratio model of 14
Minnesota lakes extracted from Frame 1345-16322
(3 July 1973) 213
D9-2 IR2 concatenation of 14 Minnesota lakes extracted
from Frame 1345-16322 (3 July 1973) 214
D10-1 Three-dimensional MSS color ratio model of 8
Minnesota lakes extracted from Frame 1346-16381
(4 July 1973) 218
D10-2 IR2 concatenation of 8 Minnesota lakes extracted
from Frame 1346-16381 (4 July 1973) 219
Dll-1 Three-dimensional MSS color ratio model of 7 New
York lakes extracted from Frame 1027-15233
(19 August 1972) 224
Dll-2 IR2 concatenation of 7 New York lakes extracted
from Frame 1027-15233 (19 August 1972) 225
D12-1 Three-dimensional MSS color ratio model of 5 New
York lakes extracted from Frame 1080-15180
(11 October 1972) 229
D12-2 IR2 concatenation of 5 New York lakes extracted
from Frame 1080-15180 (11 October 1972) 230
-------
LIST OF TABLES
Table Page
1 LANDSAT-1 orbital parameters 6
2 LANDSAT-1 MSS characteristics 9
3 Study lakes 24
4 Distribution of the world's estimated water supply 32
5 Trophic indicators and their response to increased 36
eutrophication 36
6 Descriptive statistics of 100 lakes 40
7 R-mode correlation matrix of six trophic state
indicators 47
8 Normalized eigenvectors and eigenvalues 49
9 Product-moment correlation coefficients of the
trophic state indicators and the principal com-
ponents 50
10 Principal component ordination and mean composite
rank ordination of 100 lakes 51
11 Descriptive statistics of Lake Koshkonong MSS
data extracted from Frame 1017-16093 CCTs 63
12 Optical properties of pure water 73
13 Indices commonly used to assess eutrophication 83
*»
14 LANDSAT-1 MSS frames 86
15 Dates of LANDSAT-1 coverage 87
16 Area! aspects of 20 NES-sampled lakes extracted
from LANDSAT-1 MSS Frames 1017-16091 and 1017-16093 91
17 Correlations between ground truth and LANDSAT-1 MSS
data (colors and color ratios) for 20 lakes in
Frames 1017-16091 and 1017-16093 92
18 Analysis of variance table of a regression model for
the prediction of Secchi disc transparency 94
-------
Table page
19 Seccht disc transparency residuals 94
20 Analysts of variance table of a regression model
for the prediction of chlorophyll a levels 97
21 Chlorophyll a_ residuals 97
22 Correlations between LANDSAT-1 MSS data
(colors and color ratios) collected on three
dates and the trophic status of 20 Wisconsin
lakes 99
23 Analysis of variance table of a regression model
for the prediction of the trophic status of lakes
found in LANDSAT-1 MSS Frames 1017-16091 and
1017-16093 (9 August 1972) 100
24 Trophic state index (PCI) residuals of 20
Wisconsin lakes found in LANDSAT-1 MSS Frames
1017-16091 and 1017-16093 (9 August 1972) 101
25 Analysis of variance table of a regression
model for the prediction of the trophic status
of 20 Wisconsin lakes found in LANDSAT-1 MSS
Frames 1323-16194 and 1323-16100 (11 June 1973) 102
26 Trophic state index (PCI) residuals of 20
Wisconsin lakes found in LANDSAT-1 MSS Frames
1323-16194 and 1323-16100 (11 June 1973) 102
27 Analysis of variance table of a regression model
for the prediction of the trophic status of 20
Wisconsin lakes found in LANDSAT-1 MSS Frames
1359-16091 and 1359-16094 (17 July 1973) 103
28 Trophic state index (PCI) residuals of 20
Wisconsin lakes found in LANDSAT-1 MSS Frames
1359-16091 and 1359-16094 (17 July 1973) 103
29 Analysis of variance table of a regression model
for the prediction of trophic status of 20
Wisconsin lakes using MSS color ratios from
three dates (9 August 1972, 11 June 1973, and
17 July 1973) 105
XT. IT
-------
Table ^i
3Q Trophic state index (PCI) residuals of 20
Wfscons-tn lakes from a regress ton model
Incorporating mean MSS color ratios from
three dates (9 August 1972, 11 June 1973,
and 17 July 1973) 105
31 Regression models developed for the estimation
of trophic state using LANDSAT-1 MSS data 106
32 Lake trophic state index class assignments for
the ADP technique 116
33 ADP results for 9 August 1972 using a 19 class
classification 117
Dl-1 MSS descriptive statistics for 15 Wisconsin
lakes extracted from Frame 1017-16093
(9 August 1972) 164
Dl-2 MSS descriptive statistics for 5 Wisconsin lakes
extracted from Frame 1017-16091 (9 August 1972) 165
D2-1 Areal aspects of 6 Wisconsin lakes extracted from
Frame 1036-16152 (28 August 1972) 168
D2-2 MSS descriptive statistics for 6 Wisconsin lakes
extracted from Frame 1036-16152 (28 August 1972) 169
D3-1 Areal aspects of 23 Wisconsin lakes extracted
from Frames 1323-16094 and 1323-16100 (11 June 1973) 173
D3-2 MSS descriptive statistics for 23 Wisconsin lakes
extracted from Frame 1323-16100 (11 June 1973) 174
D4-1 Areal aspects of 21 Wisconsin lakes extracted from
Frames 1359-16091 and 1359-16094 (17 July 1973) 178
D4-2 MSS descriptive statistics for 4 Wisconsin lakes
extracted from Frame 1359-16091 (17 July 1973) 179
D4-3 MSS descriptive statistics for 17 Wisconsin lakes
extracted from Frame 1359-16094 (17 July 1973) 180
D5-1 Correlations between ground truth and MSS data
(colors and color ratios) for 11 Minnesota lakes
in Frame 1022-16373 (14 August 1972) 181
xiv
-------
Table Page
D5-2 Analysts of variance table for PCI regression
model for 11 Minnesota lakes tn Frame 1022-16373
(14 August 1972) 182
D5-3 PCI restduals of 11 Minnesota lakes in Frame
1022-16373 (14 August 1972) 182
D5-4 Analysis of variance table of the Secchi disc
transparency regression model of 11 Minnesota
lakes in Frame 1022-16373 (14 August 1972) 183
D5-5 Secchi disc transparency residuals of 11
Minnesota lakes in Frame 1022-16373
(14 August 1972) 183
D5-6 Area! aspects of 12 Minnesota lakes extracted
from Frame 1022-16373 (14 August 1972) 187
D5-7 MSS descriptive statistics for 12 Minnesota lakes
extracted from Frame 1022-16373 (14 August 1972) 188
D6-1 Correlations between ground truth and MSS data
(color ratios) for 12 Minnesota lakes in Frame
1075-16321 (6 October 1972) 189
D6-2 Analysis of variance table for the PCI regression
model for 12 Minnesota lakes extracted from Frame
1075-16321 (6 October 1972) 190
D6-3 PCI residuals of 12 Minnesota lakes extracted from
Frame 1075-16321 (6 October 1972) 190
D6-4 Areal aspects of 15 Minnesota lakes extracted from
Frame 1075-16321 (6 October 1972) 195
D6-5 MSS descriptive statistics for 15 Minnesota lakes
extracted from Frame 1075-16321 (6 October 1972) 196
D7-1 Correlations between ground truth and MSS data
(color ratios) for 7 Minnesota lakes extracted
from Frame 1077-16431 (8 October 1972) 197
D7-2 Analysis of variance table of the PCI regression
model for 7 Minnesota lakes extracted from Frame
1077-16431 (8 October 1972) 198
xv
-------
Table
Page
D7-3 PCI residuals of 7 Minnesota lakes extracted from
Frame 1077-16431 (8 October 1972) 198
D7-4 Areal aspects of 10 Minnesota lakes extracted from
Frame 1077-16431 (8 October 1972) 202
D7-5 MSS descriptive statistics for 10 Minnesota lakes
extracted from Frame 1077-16431 (8 October 1972) 203
D8-1 Correlations between MSS data (colors and color
ratios) and PCI values for 11 Minnesota lakes
extracted from Frame 1309-16325 (28 May 1973) 204
D8-2 Analysis of variance table of the PCI regression
model for 11 Minnesota lakes extracted from Frame
1309-16325 (28 May 1973) 205
D8-3 PCI residuals of 11 Minnesota lakes extracted
from Frame 1309-16325 (28 May 1973) 205
D8-4 Descriptive statistics for 13 Minnesota lakes
extracted from Frame 1309-16325 (28 May 1973) 206
D8-5 Areal aspects of 13 Minnesota lakes extracted
from Frame 1309-16325 (28 May 1973) 209
D9-1 Correlations between MSS data (color ratios) and
PCI values for 12 Minnesota lakes extracted from
Frame 1345-16322 (3 July 1973) 210
D9-2 Analysis of variance table of the PCI regression
model for 12 Minnesota lakes extracted from Frame
1345-16322 (3 July 1973) 211
D9-3 PCI residuals of 12 Minnesota lakes extracted from
Frame 1345-16322 (3 July 1973) 211
D9_4 MSS descriptive statistics for 14 Minnesota lakes
extracted from Frame 1345-16322 (3 July 1973) 212
D9-5 Areal aspects of 14 Minnesota lakes extracted from
Frame 1345-16322 (3 July 1973) 215
D10-1 Correlations between MSS data (colors and color
ratios) from the PCI values for 7 Minnesota lakes
extracted from Frame 1346-16381 (4 July 1973) 216
xvi
-------
Tab!e page
D1Q-2 Analysts of variance table of the PCI regression
model for 7 Minnesota lakes extracted from Frame
1346-16381 (4 July 1973) 217
D10-3 PCI residuals of 7 Minnesota lakes extracted from
Frame 1346-16381 (4 July 1973) 217
D10-4 Areal aspects of 8 Minnesota lakes extracted from
Frame 1346-16381 (4 July 1973) 220
D10-5 MSS descriptive statistics for 8 Minnesota lakes
extracted from Frame 1346-16381 (4 July 1973 221
Dll-1 Correlations between ground truth and MSS data
(colors and color ratios) for 7 New York lakes extracted
from Frame 1027-15233 (19 August 1972) 222
Dll-2 Analysis of variance table of the PCI regression
model for 7 New York lakes extracted from Frame
1027-15233 (19 August 1972) 223
Dll-3 PCI residuals of 7 New York lakes extracted from
Frame 1027-15233 (19 August 1972) 223
Dll-4 Areal aspects of 7 New York lakes extracted from
Frame 1027-15233 (19 August 1972) 226
Dll-5 MSS descriptive statistics for 7 New York lakes
extracted from Frame 1027-15233 (19 August 1972) 227
D12-1 Areal aspects of 5 New York lakes extracted from
Frame 1080-15180 (11 October 1972) 231
D12-2 MSS descriptive statistics for 5 New York lakes
extracted from Frame 1080-15180 (11 October 1972) 232
xv n
-------
ACKNOWLEDGEMENTS
The scale and scope of this research effort was made possible through
the cooperation of the U.S. Environmental Protection Agency, the National
Aeronautics and Space Administration, the Jet Propulsion Laboratory
(California Institute of Technology), and Oregon State University.
Sincere thanks are extended to Dr. Jack H. Gakstatter of EPA's National
Eutrophication Survey for providing ready,access to unpublished lake
survey data and the welcomed funding for the project. James M. Omernik
focused the author's attention on the gray tone differences among NES-
sampled lakes found on LANDSAT-1 MSS photographs and suggested that the
differences might relate to variations in water quality. Dr. Norbert A.
Jaworski is acknowledged for providing the administrative support so
vital to the project's completion. Vernard H. Webb of EPA's Environ-
mental Photographic Interpretation Center (EPIC) ably filled the posi-
tion of NASA principal investigator and facilitated the procurement of
the MSS computer-compatible tapes.
Kenneth V. Byram, Lyle A. Wilson, and William C. Tiffany supplied com-
puter programming assistance. Additional logistic support was provided
by: Marvin 0. All urn, Albert Katko, Cecillia 0. Boland, Clarence A".
Callahan, Ralph E. Austin, Karen Randolph, and Betty McCauley. Dr.
Leslie P. Seyb, Richard J. Blackwell, and Dr. D. Phillips Larsen are
thanked for their review commentaries. The manuscript was typed by
Carla R. Juilfs, Beverley P. Bowman, and Donald A. Belshee.
It is a pleasure to acknowledge the crucial assistance provided by Dr.
Richard H. Green and the personnel of JPL's Image Processing Laboratory.
Richard J. Blackwell's expertise in digital image processing was inval-
uable; his contribution in uncountable hours of concerned discussion
and involvement is far greater than the few explicit citations indicate.
Appreciation is extended to Dr. Robert E. Frenkel, Dr. Robert C. Bard,
Dr. J. Granville Jensen, Dr. Larry G. Forslund, Dr. Henry A. Froehlich,
and Dr. Donald A. Pierce, faculty members at Oregon State University.
Dr. Frenkel served as the author's major advisor. Dr. Pierce offered
many valuable suggestions relating to statistical techniques.
The Environmental Remote Sensing Applications Laboratory at Oregon
State University is acknowledged for permitting the investigator the
use of its microdensitometer and for providing access to its LANDSAT-1
browse film library.
This research effort was supported under NASA Contract Number NAS7-100
to the Jet Propulsion Laboratory and JPL Contract Number EZ-608770.
Additional computer time was provided through the Oregon State Univer-
sity Computer Center and the National Eutrophication Survey Program.
xvi i i
-------
SECTION I
CONCLUSIONS
Based on the ground truth arid the milltiSpectral seamier frames selected
for analysis, the image extraction and enhancement techniques and data
reduction methods employed, it is conclude^ that:
1. The LANDSAT-1 MSS is an effective tool for lake enumeration and for
the estimation of lake surface area.
2. Estimates of lake Secchi disc transparency and ehlorophyll a_ levels
having practical significance can be achieved through the incorpora-
tion of lake MSS color ratios in regression models. Each trophic
indicator has a model which 1s unique to the specific date of
LANDSAT-1 coverage.
3. MSS color ratios can be used to estimate lake position on a multi-
variate trophic scale. However, each date of LANDSAT-1 coverage
has its unique model- The models for different dates vary greatly
in their predictive Capabilities. The standard error of the models
tends to decrease as the growing season progresses and the mani-
festations of eutrophication become more evident.
4. While information relating to lacustrine trophic state can be ex-
tracted from LANDSAT-1 MSS imagery, either by visual inspection or
through nricrodensitorrietry and optical density slicing, the maximum
benefits in water-based studies can be derived only through the use
of the digital data contained on the computer-compatible tapes in
conjunction with automatic image processing techniques*
5. Although LANDSAT-1 provides 18-day repetitive coverage, systematic
times-series are difficult, if not impossible, to obtain due to
excessive cloud cover on many dates of satellite coverage.
6. The LANDSAT-1 MSS has utility as a supplemental data source in lake
survey and monitoring programs. Its value is most apparent in
situations involving large lakes and/or large numbers of lakes.
-------
SECTION II
RECOMMENDATIONS
In January 1975, NASA inserted its second Earth Resources Technology
Satellite (ERTS-2, i.e., LANDSAT-2) into a sun-synchronous near-polar or-
bit. The orbital parameters of LANDSAT-2 and its still operational pre-
decessor, LANDSAT-1, have resulted in 9-day repetitive coverage. The
tandem combination of satellites has effectively doubled the amount of
imagery and digital data available to investigators. The increased cover-
age should be of great value in lake monitoring and survey programs.
Lakes are by their very nature dynamic and the more frequent satellite
coverage should increase the accuracy of the trophic state estimates.
NASA has been authorized to launch a third Earth Resources Technology
Satellite (ERTS-3, i.e., LANDSAT-3) into orbit; the tentative launch date
is February 1977. The instrumentation will include an improved multi-
spectral scanner having greater spatial resolution and capabilities in
the thermal region of the spectrum. It is anticipated that the improve-
ments will increase the utility of the satellite-borne sensor as a means
of classifying lakes.
As of September 1975, the LANDSAT series of earth resources satellites has
amassed an inventory numbering in excess of 150,000 frames (scenes) of the
United States. The collection is expanding at an average rate of about
190 frames per day. It is maintained by the EROS Data Center at Sioux
Falls, South Dakota.
A magnetic tape library has been established at JPL's Image Processing
Laboratory for the lakes examined during the course of this investigation.
The lake images have been extracted from their terrestrial matrix and are
readily available for additional statistical analyses using automatic
digital image processing techniques. The enlargement of this library
through continued acquisition of MSS data for NES-sampled lakes will give
researchers the opportunity to study long-term changes in the lakes.
In light of the preceding information and the conclusions drawn from this
investigation, it is recommended that additional effort be expended to
refine the predictive capabilities of the LANDSAT MSS. Several problem
areas which merit additional consideration have been identified.
The development of a more representative trophic state index may well be
in order. . Although automatic data processing techniques (such as cluster
analysis) have a role in the classification of lakes, the kinds and num-
ber of trophic indicators, the types of data transformations, and the
methodologies employed can produce significant differences in the
-------
resulting classifications, Very little work has been done in this area
using the techniques commonly employed in numerical taxonomy. ADP tech-
niques are well suited for the reduction of the large data masses gener-
ated by lake survey programs.
Although atmospheric and solar angle effects were disregarded, it is ap-
parent that they can degrade the lake color signal received by the multi-
spectral scanner. The development of radiometric calibration techniques
to account for the peripheral effects would greatly increase the utility
of the MSS. An increase in instrument gain (this is possible in the RED
and GRN bands) would aid the water-based investigator, possibly at the
expense of terrestrial studies.
The need exists to study in greater detail the relationships between the
magnitude of trophic state indicators and MSS data under both laboratory
and field conditions. This study suffered in part from the lack of con-
current data; it is imperative that concurrent satellite-ground truth
sampling be conducted to further elucidate the relationships.
Bottom effects were assumed to be insignificant in this study because
Secchi disc transparency was much less than water depth. However, in
some lakes the bottom can be seen by the sensor. A better knowledge of
bottom effects may be of value in refining the predictive capabilities of
the MSS in the aquatic environment.
With the exception of some lakes in the New York frames, the remote sen-
sing aspects of this investigation have been heavily weighted with lakes
that are often referred to as "eutrophic". The need exists to examine
more lakes, particularly those possessing high water quality. The sensi-
tivity limits of the MSS have yet to be established for trophic state in-
dicators as well as the trophic state index. The examination of lakes
from other geographic regions should be undertaken to determine whether
the relationships found in the study area extend to other regions.
-------
SECTION III
INTRODUCTION
Homo sapiens has developed the burgeoning population and the techno-
logical level necessary to effect significant alterations in the
biosphere. His impact on environment is accelerating in variety,
magnitude, and geographic extent. He has failed, either through
gross ignorance or a lack of concern, to give adequate consideration
to the interrelatedness of the biotic and abiotic elements which
compose the environment. Man-induced environmental alterations
generally result in a chain of consequences, both direct and indirect,
short-term and long-term, and varying in magnitude. Many of the
consequences (e_.c[_., air pollution, water pollution) adversely affect
his health, his economic well-being, and the pristine qualities of
environment.
Man's strategies for using the earth's lacustrine resources have
usually been predicated on immediate short-term economic gains with
little consideration of long range environmental ramifications. It is
becoming increasingly apparent that man has adversely affected many
lakes, particularly those located in countries with large population
densities and/or which are technologically advanced. The United States
of America can serve as a prime example.
STATEMENT OF PURPOSE
Rational management of the lacustrine resource dictates, as the first
step, an assessment of each lentic body's trophic status. Data col-
lection for the determination of trophic status is a costly, time-
consuming process, especially when thousands of lakes are to be eval-
uated. The need exists to find a means of rapidly assessing the trophic
state of water bodies which would make it economically feasible to
operate extensive systematic surveillance programs. Stewart and
Rohlich (1967) have urged investigators to develop remote sensing
techniques and evaluate their potential for eutrophication surveillance.
Satellite-borne sensors show promise as a means of monitoring and
classifying lakes and reservoirs. The successful orbiting of the
Earth Resources Techology Satellite (ERTS-1; i.e.., LANDSAT-1) affords
the opportunity to investigate the potential of one type of satellite-
borne sensor.
The purpose of this report is the evaluation of LANDSAT-1 remotely
sensed multispectral scanner data as a means of determining the
-------
trophic state of a selected group of lakes located in the northern part
of the conterminous United States. Specific objectives include: the
development and application of a multivariate trophic ranking system to a
selected group of lakes; the formulation of empirical models for the
estimation of selected trophic state indicators; and the development of
empirical models for the prediction of lacustrine trophic state using
LANDSAT-1 data.
FORMAT OF REPORT
The remainder of this section is devoted to a description of the
Earth Resources Technology Satellite, its orbital characteristics,
instrumentation, and products. In addition, the study area is described
along with the criteria used in the selection of the lakes and the tech-
niques used in the collection of the ground (I.e.., water) truth.
Section IV discusses lakes as natural resources, the concept of
eutrophication, trophic state indicators, and demonstrates the use
of multivariate techniques in the classification of a large group of
lakes on the basis of selected ground truth parameters.
Section V is devoted to a detailed description of the methodology used
in the extraction and transformation of LANDSAT-1 multispectral scanner
(MSS) data into forms and products which can then be used in the study
of hypothesized ground truth-MSS data relationships.
Correlations between lake parameters (surface area, chlorophyll a.,
Secchi disc transparency) and MSS data are explored in Section VI.
Regression models are developed to predict the magnitude of selected
trophic state indicators.
Section VII includes the prediction of lake trophic state using MSS
color ratios in regression models. Three-dimensional models are used
to illustrate qualitative differences among the lakes on the basis of
MSS color ratios. An automatic image processing technique is employed
to generate enhanced photographic products depicting the trophic state
of selected lakes.
Section VIII is devoted to a brief discussion and summation of the
potential applications and limitations of the LANDSAT-1 MSS in lake
monitoring and classification.
THE EARTH RESOURCES TECHNOLOGY SATELLITE
The Earth Resources Technology Satellite Program, under the sponsorship
of the National Aeronautics and Space Administration, is a concerted
effort to merge space and remote sensing technologies into a system
which will demonstrate techniques for efficient management of the
earth's natural resources. To explore the feasibility of applying
earth resource data collected from satellite altitudes to resource
-------
management problems, NASA inserted an experimental satellite (ERTS-A)
into a circular earth orbit. Another satellite (ERTS-B) is scheduled
for launch when ERTS-A, officially designated ERTS-1 (i.e.., LANDSAT-1)
ceases to function. Detailed information regarding the Earth Resources
Technology Program is found in the Data Users Handbook (NASA, 1972).
ERTS-1 Orbit Parameters and Earth Coverage
LANDSAT-1 was placed into a nominal sun-synchronous near-polar orbit by
a Delta launch vehicle on 23 July 1972 (Freden, 1973). Orbital param-
eters are listed in Table 1.
TABLE 1. LANDSAT-1 ORBITAL PARAMETERS3
Orbit Parameter Actual Orbit
Semi-major axis 7285.82 kilometers
Inclination 99.114 degrees
Period 103.267 minutes
Eccentricity 0.0006
Time at descending node 9:42 a.m.
(southbound equatorial crossing)
Coverage cycle duration 18 days
(251 revolutions)
Distance between adjacent ground ,159.38 kilometers
tracks (at equator)
aAdapted from Data Users Handbook (NASA, 1972).
The earth coverage pattern is shown in Figure 1 for two orbits on two
consecutive days. Orbital parameters result in a 1.43 degree westward
migration of the daily coverage swath, equivalent to a distance of 159
kilometers at the equator. The westward progression of the satellite
revolutions continues, exposing all of the area between orbit N and
orbit N+l to the satellite sensors by day M.. This constitutes one
complete coverage cycle and consists of 251 revolutions. The cycle
takes exactly 18 days and results in total global coverage between 81°N
and 81°S latitude. Fourteen orbits (i_.e_., revolutions) are completed
during each of 17 days in a cycle with 13 revolutions during one day
(NASA, 1972). Approximately 188 scenes are acquired on an average day
(Nordberg, 1972).
-------
ORBIT N + 1, DAY M + 1
ORBIT N + 1, DAY M
ORBIT N, DAY M + 1
ORBIT N,DAY M
NOTE: ORBIT N, DAY M + 1
OCCURS 14 REVOLUTIONS
AFTER ORBIT N, DAY M
EQUATOR = 159 km
Figure 1. LANDSAT-1 ground coverage pattern. Adapted from Data Users Handbook (NASA, 1972)
-------
LANDSAT-1 Instrumentation
The LANDSAT-1 payload consists of a Return Beam Vidicon (RBV) Camera
Subsystem, a Multispectral Scanner Subsystem (MSS), and a Data
Collection System (DCS). The RBV and MSS are designed to provide
multispectral imagery of the earth beneath the observatory (i.e.,
satellite). A malfunction occurred on 6 August 1972 (orbit 198) in
the RBV power switching circuit and the RBV cameras were turned off as
only one sensor system can be used in conjunction with the one function-
ing video tape recorder. The second recorder aboard the observatory
malfunctioned between orbits 148 and 181 (Freden, 1973). The DCS serves
to relay environmental information from geographically remote ground-
based sensors to LANDSAT ground stations for processing and delivery to
users. The RBV and DCS aspects of the satellite need not concern us.
The MSS is a line-scanning radiometer which collects data by creating
images of the earth's surface in four spectral bands simultaneously
through the same optical system. The instrument operates in the solar-
reflected spectral band region from 500 to 1,100 nanometers. Scanner
characteristics are listed in Table 2. The MSS scans cross-track swaths
185 kilometers in width, simultaneously imaging six scan lines for each
of the four bands. The object plane is scanned by an oscillating flat
mirror positioned between the scene and a double reflector telescope-
type of optical chain. An 11.5 degree (Horan, Schwartz, and Love, 1974)
cross-track field of view is produced by the mirror oscillating ±2.89
degrees about its nominal position (Figure 2).
A nominal orbital velocity of 6.47 kilometers per second, neglecting
observatory perturbations and earth rotation effects, produces the
requisite along-track scan. The subsatellite point moves 474 meters
along the track during the 73.42 millisecond active scan-retrace
cycle which is itself a consequence of the 13.62 hz mirror oscillation
rate. The track distance of 474 meters synchronizes with the 474 meter
along-track field of view of each set of six detectors. The line
scanned by the first detector in one cycle of the active scan is in
juxtaposition to the line scanned by the sixth detector of the previous
scan (Figure 3).
Twenty-four glass optical fibers, arranged in a four by six (4x6) matrix
in the focused area of the telescope, intercept the light from the
earth scene. Light impinging on the square input end of each optical
fiber is conducted to an individual detector through an optical filter
unique to the respective spectral band under consideration. Photo-
multiplier tubes (PMT) serve as detectors for Bands Four through Six;
Band Seven uses silicon photodiodes. A video signal is produced at the
scanner electronics output as the image of a line across the swath is
swept across the fiber during active scan. The signal is sampled at
9.95 microsecond (ysec) intervals which correspond to a 56 meter cross-
track motion of the instantaneous field of view. The sampled signal is
digitized and arranged into a serial digit data stream for transmission
8
-------
TABLE 2. LANDSAT-1 MSS CHARACTERISTICS0
Item
Characteristics
Telescope optics
Scanning method
Scan (Swath) width
Scan duty cycle
Instantaneous field of view
(IFOV)
Number of bands
Number of lines (detectors)
scanned per band
Limiting ground resolution from
917 kilometers altitude
Spectral band wavelength:
NDPF Band Code
Band 4 (Green)
Band 5 (Red)
Band 6 (Near-infrared One)
Band 7 (Near-infrared Two)
Sensor response:
Detector
Nominal input for 4 V Scanner
Output (10'HJ cm ~2sr-1)
Scanner and multiplexer weight
Signal channels
Telemetry channels
Command capability
Scanner size
22 cm (aperture diameter),
f/3.6 Ritchey-Chretien
Flat mirror oscillating ±2.9
degrees at 13.62 Hz
11.5 degrees (185 kilometers at
917 kilometers altitude)
44%
86 microradians
Four
Six
40 meters
500-600 nanometers
600-700 nanometers
700-800 nanometers
800-1,100 nanometers
Band 4 Band 5 Band 6
Band 7
PMT°
24.8
PMT
20.0
PMT
17.6
Photodiode
46.0
50 kilograms
24
97
72
Approximately 36 x 38 x 89 cm
Adapted from Horan, e_t al_. (1974)
Dphotomultiplier tube.
-------
SCANNER
6 DETECTORS
PER BAND:
24 TOTAL,
+2 FOR BAND 5
(ERTS B)
^—OPTICS
\
^:
C_I£
/
V"
SCAN MIRROR
(OSCILLATES
NOMINALLY
+ 2.89°)
NOTE: ACTIVE SCAN IS
WEST TO EAST.
FIELD OF VIEW
= 11.56 DEGREES
185 km
(100 ran)
PATH OF
SPACECRAFT
TRAVEL
6 LINESACAN/BAND
Figure 2. Schematic diagram of the LANDSAT-1 MSS scanning
arrangement. Adapted from the Data Users Handbook
(NASA, 1972).
10
-------
£.
bo1
ex
o
at
nf
X
*~c
I
O
ex.
tx.
i —
Co
LINE 1
2
3
4
5
6
LINE 1
2
3
4
5
6
SPACECRAFT
VELOCITY VECTOR
185 km
WIDTH
COMPOSITE
TOTAL AREA SCAN
FOR ANY BAND
FORMED BY
REPEATED 6 LINE
PER BAND SWEEPS
PER ACTIVE
MIRROR CYCLE
Figure 3. Ground scan pattern for a single MSS detector. Adapted
from the Data Users Handbook (NASA, 1972).
-------
to ground stations. Individual signals are derived from light passing
through each fiber, resulting in 24 channels of output.
LANDSAT-1 Products
The electronic signals from the observatory's MSS are converted into
photographic and computer products at the Goddard Space Flight Center,
Greenbelt, Maryland. Third and fourth generation photographic products
are available in the form of prints, positive and negative transparen-
cies, and come in several scales including: 1:3,369,000; 1:1,000,000;
1:500,000; and 1:250,000 with the transparencies being limited to the
two smaller scales. Color products are available for a relatively
small number of scenes. Computer-compatible magnetic digital tapes
(CCT's) may be requested in either a 7-track or a 9-track format. Four
CCT's are required for the MSS data corresponding to one scene.
Copies of the CCT's and the photographic products are placed in the
public domain at the Department of the Interior's Earth Resources
Observations Systems (EROS) Data Center located 1n Sioux Falls,
South Dakota.
LANDSAT-1 Lake Monitoring Potential
The advantages of using remote sensing imagery systems are threefold:
they afford an overall (synoptic) view, they can give a time record,
and they expand the spectral limits of the human eye (Scherz, et al., 1969;
Scherz, 1971a; Scherz, 1971b). Multispectral satellite-borne imagery
systems show promise as a means of monitoring and classifying the
earth's lacustrine resources. This is partially due to the repetitive
nature of a satellite and the tremendous synoptic view offered from
orbital altitudes.
Visual examination of select frames of LANDSAT-1 MSS imagery from
Wisconsin, Minnesota, and Florida suggest that good correlations may
exist between the trophic status of lakes and their tonal characteris-
tics. MSS Frame 1017-16093, recorded at an altitude of approximately
917 kilometers over southeastern Wisconsin and northeastern Illinois on
9 August 1972, will serve to illustrate this point.
Figure 4 is a reproduction of an EROS Data Center photograph of the
scene as recorded in the near-infrared (IR2; 800 to 1,100 nanometers)
spectral band. Water bodies, including the larger streams, stand out
boldly against the lighter tones of the land features. The labelled
lakes, excluding Lake Michigan, were sampled by the U.S. Environmental
Protection Agency's National Eutrophlcation Survey (NES) during the
1972 open water season. Gray tone differences are not evident among
the lakes nor are tonal patterns visible on any of the lakes. IR2 is,
however, a good spectral band for the location and demarcation of water
bodies. Some caution is necessary when conducting a lake enumeration
12
-------
on the photograph because some of the "lakes" are in reality shadows
cast by cumulus clouds.
Figure 5 is the same scene recorded in near-infrared one (IR1; 700 to
800 nanometers). Tonal differences are apparent, at least in the
original photographic print, among the lakes, and patterns are
evident on some of the lakes (e_.g_., Lake Koshkonong). Lakes are
readily located and their boundaries delimited in this band.
Figure 6 is a red light (RED; 600 to 700 nanometers) MSS photograph of
the scene. Marked gray tone differences are apparent among the lakes.
Lakes commonly recognized as eutrophic (e_.g_., Lake Como) tend to appear
light in tone and meld in with the land features. Lakes with relatively
good water quality (e_.g_., Lake Geneva) are characterized by darker tones.
Lotic bodies are not readily apparent in the photograph.
The green light (GRN; 500 to 600 nanometers) sensed by the MSS was used
to construct the Figure 7 photograph. Although the lakes are difficult
to discern, a result of low contrast among the scene elements, differ-
ences among the lakes can be detected with the unaided eye.
It is apparent, from the visual examination of LANDSAT-1 MSS Frame
1017-16093 and other frames from several additional states, that the
satellite-borne multispectral scanner is collecting data which may be
of value in the classification and monitoring of lentic bodies. The
results of the examination suggest that GRN, RED, and IR1 contain most
of the information relative to trophic status assessment.
DESCRIPTION OF THE STUDY AREA
Many current efforts to determine the feasibility of using LANDSAT-1
MSS data in water quality monitoring tend to be intensive in nature,
near-real time, and oriented toward the dymanics of pollution or
eutrophication (e_._g_., Chase and Smith, 1973; Lind and Henson, 1973;
Lind, 1973). The investigations are very limited in geographic scope,
typically involving fewer than six water bodies.
Another approach, the one used in this report, involves the examination
of a relatively large number of lakes over a more extensive geographic
area. Although it is not uncommon to find lakes of different trophic
state within the same lake region (e_.g_., Southeastern Lake District of
Wisconsin), the selection of a larger study area affords the opportuni-
ty to include lakes exhibiting a greater trophic range. In addition,
assuming that lakes have characteristic multispectral signatures at
particular points on the trophic scale, the use of a larger and more
diverse lake population permits a more extensive evaluation of the
LANDSAT-1 MSS's capabilities in the realm of lake monitoring and
classification.
13
-------
n*3-w*oii-yi N mj-y>'t«iut-M "«»„_JLJLHB
HMB-381
Figure 4. Reproduction of an EROS Data Center IR2 print of Frame
1017-16093 (9 August 1972). Water bodies stand out in
stark contrast to the lighter-toned land features. The
labelled lakes, excluding Lake Michigan, were sampled
by the National Eutrophication Survey during 1972. The
reproduction, originally printed at a scale of 1:1,000,000,
has a scale of approximately 1:1,415,000.
14
-------
99HUG72 C N«3
• H-C *
-•'.• • NAGAWICKA a
U088-80I
MSS 6 D SJN EL5" BZI3B 192-8236-G-1-N-D-2L KflSB ERTS E 1817- I6B93-6 81
Figure 5. Reproduction of EROS IR1 print of Frame 1017-16093
(9 August 1972). Surface patterns are evident on some
of the lakes (e.g., Lake Koshkonong). Each edge of the
picture is equivalent to a ground distance of approxi-
mately 185 kilometers. The reproduction is printed at
a scale approximating 1:1,415,000.
15
-------
, •
Figure 6. Reproduction of EROS RED (MSS Band 5) print of Frame 1017-
16093 (9 August 1972). Variations in gray tone are readily
apparent among the lakes and suggest differences in water
quality. Lake Geneva, characterized by relatively high water
quality, is dark in tone compared with, for example, eutro-
phic Lake Koshkonong. The small ball-like white objects
between Milwaukee and Chicago are cumulus clouds. The
reproduction is printed at a scale of approximately 1:1 415 _
000.
16
-------
B9BUG72 C N*3
O
•a
a
e
.' J-I6893-* 81
Figure 7. Reproduction of EROS GRN (MSS Band 4) print of Frame 1017-
16093 (9 August 1972). Scenes recorded in the green band
generally lack contrast, but contain information useful in
monitoring earth resources. The reproduction is printed
at a scale of 1:1,415,000.
17
-------
Geographic Area
The geographic area serving as a matrix for the study lakes comprises
the states of Minnesota, Wisconsin, Michigan, and New York. The area
was selected for its abundance of lakes, the availability of pertinent
environmental (i_.e_., ground truth) data and concurrent or near-
concurrent LANDSAT-1 MSS frames.
Climate
The climate of the study area is humid continental. Continental polar
air masses dominate during the winter, resulting in a mean temperature
of less than -10°C for the coldest month. Most of the study lakes are
ice-bound for several months, two notable exceptions being the New York
Finger Lakes, Seneca and Cayuga, which usually remain open and in full
circulation throughout the winter (Berg, 1963). Tropical air masses
dominate during the summer and high temperatures prevail. The warmest
month has a mean temperature in excess of 18°C. The temperature
extremes are less severe in the eastern portion of the study area (New
York). Although precipitation is common throughout the year, a summer
maximum exists. The mean annual precipitation ranges from a low of
approximately 50 centimeters in northwestern Minnesota to a high of
about 125 centimeters in north central New York (USGS, 1970).
Detailed descriptions of the climate are found in Trewartha (1968),
USDA (1941), Visher (1954), Thornthwaite (1948), and Strahler (1969).
Geology, Soils, and Land Use
The study area extends over three physiographic divisions and includes
seven subdivisions (Figure 8). The entire area, excluding a small
region in western Wisconsin and southeastern Minnesota (the "Driftless
Area"), was exposed to the forces of continental glaciation which were
instrumental in the formation of the numerous lake basins.
The bedrock ranges in age from Precambrian to Cretaceous and includes
sedimentary, metamorphic, and igneous rocks (Figure 9). Areas underlain
with older Precambrian rocks and covered by a thin veil of glacial
drift (e_.£., northeastern Minnesota) generally have lakes possessing
relatively high water quality. The landscape is dominanted by the
members of six soil orders (7th Approximation) including Alfisols,
Entisols, Histosols, Inceptisols, Mollisols, and Spodosols (Figure 10).
Major types of land use are shown in Figure 11. Lakes seriously
modified by man are generally found in areas where land use is chiefly
agricultural in nature, particularly as cropland.
Lake Selection Criteria and Location
Most of the lakes'incorporated into this study were selected from some
220 lakes sampled in 1972 by the U.S. Environmental Protection Agency's
(EPA) National Eutrophication Survey (NES). The lakes were selected,
18
-------
PHYSICAL SUBDIVISIONS
Figure 8. Physical subdivisions. Adapted from U.S. Geological
Survey (1970) and Hammond (1964).
-------
ro
o
Figure 9. Geology. Adapted from U.S. Geological Survey (1970).
For greater detail see U.S. Geological Survey (1965),
Geologic Map of North America, 1:5,000,000.
-------
ro
PRINCIPAL KINDS OF SOILS
Figure 10. Distribution of principal kinds of soils: Orders, Suborders,
and Great Groups. The letter-number symbols in the legend are
abbreviated from those on the map. For a complete description,
see U.S. Geological Survey (1970). Complete definitions of the
soils are found in Soil Classification, A Comprehensive System,
7th Approximation (U.S. Department of Agriculture, 1960) and
(USDA, March 1967) Supplement. The map is adapted from U.S.
Geological Survey (1970).
-------
r\3
Figure 11. Major land uses. Adapted from U.S. Geological Survey (1970).
-------
for the purposes of this report, to give extensive geographic coverage,
trophic indicator values that covered a range representative of the
1972 NES-sampled lakes, and for completeness of sampling records. The
study lakes are listed in Table 3. NES-sampled lakes have both a serial
number and a STORET (STOrage and RETrieval) number; those with only a
serial number are outside the scope of NES. Lake locations are depicted
in Figure 12.
GROUND TRUTH COLLECTION
Each NES lake (serial numbers 1-100) was sampled three times (spring,
summer, fall) during the 1972 calendar year (Appendix B) by helicopter-
borne sampling teams operating from pontoon-equipped Bell UL-1H aircraft.
The helicopters were equipped with a submersible pump and appropriate
sensors for in situ measurements of conductivity, temperature, optical
transmissivity, dissolved oxygen, hydrogen-ion concentrations (pH), and
water depth. Additional equipment included an echo sounder, 30 cm
Secchi disc, and water sampling equipment.
Lake Site Selection
Lake sampling sites were selected on the basis of lake morphometry, and
potential major sources of nutrient input, as well as the on-site
judgment of the sampling team's limnologists. The number of sampling
sites varied for different lakes, ranging from one to nine (Lake Winne-
bago).
Lake Sampling Methods
After landing on the lake surface in the general area of a sampling
site, the helicopter was taxied to locate the deepest nearby water.
There a small reference bouy was deployed to aid the pilot in maintain-
ing his station. Observations were recorded relating to general lake
appearance, phytoplankton bloom conditions, aquatic macrophytes, and
shoreline developments (e_.£., residential units) along with magnetic
compass bearings to prominent landmarks. A photograph was taken of the
site, including the reference buoy and the prominent landmarks to
assure return to the same site on subsequent sampling rounds.
Secchi disc transparency readings were made and water samples were
bucket-dipped from the lake surface. The sensor recorders were
monitored as the sensor-pump package was slowly lowered through the
water column, permitting the limnologists to select depths or levels to
be sampled as the package was winched to the surface. After touching
bottom, the package was raised to a point approximately 1.2 meters off
the bottom? Each sensor's digital output was recorded and the submer-
sible pump activated for the collection of water samples. The package
was then raised to the next sampling level and the process was repeated.
* In 1972 the sensor could be deployed to a maximum depth of about 114
meters.
23
-------
TABLE 3. STUDY LAKES
Serial Q STORET
Lake Name Numbera Number
Blackduck
Bemidji
Andrusia
Wolf
C ass
Leech
Birch
Trout
Mashkenode
Whitewater
Pelican
Shagawa
Gull
Rabbit
Cranberry
Darling
Carlos
Le Homme Dieu
Minnewaska
Nest
Green
Wagonga
Clearwater
Mud (at Maple Lake)
Cokato
Buffalo
Carrigan
Silver
Minnetonka
Forest
White Bear
St. Croix
Spring
Pepin
M ad i s o n
Sakatah
Bear
Albert Lea
Yellow
Wapogasset
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
2711
27C1
27CO
27A2
2715
2746
2710
2793
2756
2749
2765
2780
2737
2771
2720
27B4
27B9
27B5
2761
27B3
27B2
27B1
2716
2753
2719
2713
2714
2782
2760
27A9
27BO
27A7
27A6
27A4
2750
2777
2706
2702
5576
5550
County
Beltrami
Beltrami
Beltrami
Becker
Cass
Cass
Cass
Itasca
St. Louis
St. Louis
Crow Wing/
St. Louis
St. Louis
Cass
Crow Wing
Crow Wing
Douglas
Douglas
Douglas
Hennepin/
Pope
Kandiyohi
Kandiyohi
Kandiyohi
Wright/
Stearns
Wright
Wright
Wright
Wright
McLeod
Hennepin
Washington
Washington
Washington
Washington
Goodhue
Blue Earth
Le Sueur
Freeborn
Freeborn
Burnett
Polk
N.
47
47
47
47
47
47
46
47
47
47
48
47
46
46
46
45
45
45
45
45
45
45
45
45
45
45
45
44
44
45
45
44
44
44
44
44
43
43
45
45'
Lake Coordinates
Latitude W. Longitude
"
°
°
O
0
O
o
o
0
o
o
0
0
0
0
0
o
0
o
0
0
0
0
o
o
0
0
0
0
o
0
D
0
°
o
0
0
0
3
45'
29'
26'
26'
27'
15'
56'
17'
29'
29'
02'
,,55'
25'
31'
24'
56'
56'
56'
36'
16'
16'
04'
20'
13'
07'
08'
03'
53'
57'
17'
04'
46'
45'
23'
10'
14'
33i
37'
55'
19'
30"
30"
30"
30"
00"
00"
00"
00"
30"
30"
00"
00"
00"
30"
00"
00"
00"
30"
30"
00"
00"
00"
00"
30"
00"
30"
30"
00"
30"
30"
00"
00"
30"
00"
30"
00"
00"
00"
30"
30"
94°
94°
94"
94"
94»
94°
94"
93-
94»
92°
92°
91°
94"
93°
93°
95°
95°
95"
95"
95°
94°
94°
94"
93°
94°
93°
93°
94"
93°
92°
92°
92°
92°
92°
93°
93°
93"
93"
92°
92°
36'
50'
38'
40'
28'
13'
31'
25'
36'
11'
50'
53'
21'
56'
46'
23'
23'
21'
32'
56'
52'
56'
07'
59'
09'
54'
57'
13'
30'
58'
58'
491
52'
02'
49'
27'
30'
17'
25'
26'
00
00
30
30
30
00
30
00
00
00
00
00
30
00
30
00
00
30
1 I
1 f
11
11
1 f
I I
1 I
1 (
I (
1 !
1 (
i 1
1 |
1 [
1 1
1 1
1 |
1 1
00"
00
f 1
00"
30
00'
00'
30'
30'
i |
1
1
!
1
30"
00'
30'
30'
30'
00'
1
1
1
1
1
30"
00"
00'
30'
1
I
00"
301
00'
1
1
30"
24
-------
TABLE 3. STUDY LAKES (continued)
Lake Name
Long
Elk
Trout
Crystal
Tainter
Shawano
Poygan
Butte Des Morts
Winnebago
Round
Green
Swan
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac La Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
B rowns
Middle
Delavan
Como
Geneva
Charlevoix
Higgins
Houghton
Pere Marquette
White
Muskegon
Fremont
Mona
Crystal
Serial STORET
Numbex^ Number
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
5573
5575
5572
5571
5546
5539
5538
5508
5554
5566
5519
5545
5503
5520
5564
5522
5563
5532
5580
5536
5531
5557
5559
5560
5569
5513
5562
5561
2617
2695
2696
2698
2688
2659
2631
2691
2694
County
Price
Price
Vilas
Vilas
Dunn
Shawano
Winnebago
Winnebago
Winnebago/
Fond Du Lac
Waupaca
Fond Du Lac
Columbia
Dodge
Dane
Vilas
Dane/
Jefferson
Waukesha
Wauke sha
Waukesha
Waukesha
Price
Waukesha
Racine
Racine
Walworth
Walworth
Walworth
Walworth
Charlevoix
Roscommon/
Crawford
Roscommon
Mason
Muskegon/
Newaygo
Newaygo /
Muskegon
Newaygo
Muskegon
Montcalm
N.
45
45
46
46
44
44
44
44
44
44
43
45
43
42
43
42
43
43
43
43
44
43
42
42
42
42
42
42
45
44
44
43
43
43
43
43
43
Lake Coordinates'5
Latitude W. Longitude
0
0
0
0
0
0
0
0
0
0
0
°
0
0
0
0
°
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
42'
42'
04'
00'
56'
47'
06'
02'
12'
20'
51'
32'
27'
58'
04'
50'
07'
05'
06'
07'
04'
05'
46'
41'
46'
37'
37'
36'
19'
26'
24'
56'
22'
14'
26'
10'
16'
00"
00"
00"
00"
00"
30"
30"
00"
00"
00"
30"
30"
30"
00"
30"
00"
00"
30"
30"
00"
00"
00"
00"
00"
30"
00"
00"
00"
00"
00"
30"
30"
30"
00"
30"
00"
30"
90°
90°
89°
89°
91°
88°
88°
88°
88°
89°
88°
89°
88°
89°
88°
89°
88°
88°
88°
88°
88°
88°
88°
88°
88°
88°
88°
88°
85°
84°
84°
86°
86°
86°
85°
86°
84°
27'
24'
40'
36'
53'
34'
42'
34'
27'
10'
57'
23'
50'
13'
55'
01'
31'
28'
28'
23'
24'
16'
13'
14'
34'
37'
27'
26'
15'
40'
47'
27'
25'
20'
58'
17'
55'
00
30
00
39
30
00
30
00
30
00
00
30
30
30
00
30
00
30
30
30
30
00
00
30
00
30
30
00
00
30
30
00
30
00
30
30
30
] |
1 (
1 1
1 <
1 1
I |
1 1
1 !
1 1
! (
t t
1 1
1 I
[ 1
1 1
1 1
1 1
1 1
1 1
[ I
1 t
1 1
1 1
1 1
1 1
1 |
1 |
1 t
! 1
11
"
"
1 1
1 1
1 1
"
1 1
25
-------
TABLE 3. STUDY LAKES (continued)
Lake Name
Jordan
Thornapple
Strawberry
Chemung
Thompson
Ford
Union
Long
Randall
Schroon
Black
Cassadaga
Chautauqua
Conesus
Canandaigua
Keuka
Seneca
Cayuga
Owasco
Cross
Otter
Round
Saratoga
W inona
Trace
Calhoun
Big Stone
Zumbro
Oneida
Canadarago
Mendota
Monona
W aub e s a
Cottonwood
Maple
Serial STORET Count
Number a Number *
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
241
2640
2683
2699
2618
2697
2629
2685
2692
2671
3624
360Z
3607
3610
3639
3604
3617
3635
3608
3627
3611
3625
3630
3633
27A1
2792
27B6
2709
27A5
3622
3603
Z7C3
Ionia/
Barry
Barry
Livingston
Livingston
Livingston
Washtenaw
Branch/
Calhoun
Ottawa
Branch
Warren/
Essex
Jefferson/
St. Lawrence
Chautauqua
Chautauqua
Livingston
Ontario
Steuben/
Yates
Yates/
Senga/
Schuler
Cayuga/
Tompkins/
Seneca
Cayuga
Cayuga
Cayuga
Saratoga
Saratoga
Douglas
Todd
Hennepin
Big Stone
Olmsted
Oswego/
Oneida
Otsego
Dane
Dane
Dane
Lyon
Wright
N.
42
42
42
42
42
42
42
41
42
43
44
42
42
42
42
42
42
42
42
43
43
42
43
45
45
44
45
44
43
42
43
43
43
44
45
Lake Coordinates
Latitude W. Longitude;
o
°
o
o
o
o
o
o
°
0
o
0
0
0
o
o
o
o
o
o
0
o
o
o
o
o
o
o
0
°
0
o
o
o
o
45'
37'
26'
35'
37'
12'
02'
55'
00'
50'
36'
20'
06'
50'
53'
39'
56'
58'
47'
06'
09'
55'
04'
33'
50'
$6'
18'
14'
11'
50'
11'
08'
01'
36'
13'
30"
00"
30"
30"
30"
30"
30"
30"
30"
00"
00"
30"
00"
00"
30"
30"
30"
00"
00"
00"
30"
00"
30"
30"
30"
00"
00"
00"
00"
00"
00"
30"
30"
00"
30"
85
85
83
83
83
83
85
84
85
73
75
79
79
77
77
77
76
76
76
76
76
73
73
95
94
93
94
92
75
75
89
89
89
95
93
o
°
o
o
o
o
0
o
o
0
o
o
o
o
o
o
o
0
o
o
o
o
o
o
o
o
0
0
o
o
o
o
o
0
o
09'
21'
51'
51'
55'
33'
12'
20'
02'
47'
28'
19'
15'
42'
15'
09'
52'
44'
29'
25'
33'
45'
41'
23'
45'
18'
27'
29'
45'
00'
13'
16'
29'
40'
59'
00"
30"
00"
30"
30"
30"
30"
00"
30"
00"
30"
30"
00"
30"
30"
00"
00"
30"
00"
30"
00"
00"
00"
00"
30"
30"
00"
00"
30"
00"
00"
00"
30"
00"
30"
The-lakes with the serial number s 1-31, 33, 35-38, 101-103, 111, and 241 are wholly
contained in Minnesota. Lakes 32, 34, and 105, located on the borders of the state,
•are referred to as "Minnesota" lakes for convenience. Lakes 39-68 and 108-110 are
Wisconsin lakes; 69-86 are in Michigan. New York lakes have been assigned the
numbers 87-100 and 106-107.
bln the case^of seepage lakes, the geographic coordinates represent a point
on the lake surface.
26
-------
Figure 12. Location of the study lakes. The lakes with serial numbers 1-100 were used in
the cluster analyses and principal component ordination analyses.
-------
The procedure was continued until all of the selected levels were
sampled.
Water samples were collected from each selected depth for nutrient,
alkalinity, pH, conductivity, and dissolved oxygen determinations. The
samples collected for alkalinity, nitrate-nitrogen, nitrite-nitrogen,
ammonia-nitrogen, and dissolved phosphorus were preserved on-site by
the addition of 40 mg/1 of mercuric chloride. The samples were filtered
through a 0.45 micrometer membrane filter (prerinsed with de-ionized
water) and the filtrate was shipped to EPA's National Environmental
Research Center (NERC), Las Vegas, Nevada. The samples for total
phosphorus determination were preserved with a 40 mg/1 mercuric chloride
solution, but were not filtered.
Integrated samples for algae enumeration and identification purposes and
chlorophyll a_ determination were collected by raising or lowering the
package while continuing to operate the pump. The package movement was
timed to provide a water sample representative of the water column from
the surface to approximately 4.6 meters, or to a point just above the
bottom in water less than 4.6 meters in depth. The algae samples were
fixed with Lugol's solution and forwarded to NERC-Las Vegas.
A 18.9-liter water sample, composited from water collected at each
sampling depth and every sampling site (i_.je., station) during the fall
sampling period, was airmailed to the Pacific Northwest Environmental
Research Laboratory (PNERL), Corvallis, Oregon, for the determination
of its productivity potential under a set of standard conditions. The
unpreserved sample, unrefrigerated while in transit, was stored in a
freezer until the analysis could commence.
Analytical Methods
The water samples for dissolved oxygen, pH, and chlorophyll a. determina-
tions were analyzed in a mobile field laboratory at the end of each day
of sampling. The dissolved oxygen water samples, fixed and acidified
aboard the helicopters', were titrated with phenylarsine oxide in
conjunction with a starch indicator. A Beckman Field Laboratory pH
meter was used to determine the hydrogen-ion concentration of the water
samples. The samples were refrigerated until they were analyzed.
The samples for chlorophyll a^determination were refrigerated in the
dark and were analyzed using the fluorometric procedure described by
Yentch and Menzel (1963). Algal identification and enumeration included total
cell count ^Sedgewick-Rafter) and a differential count and identifica-
tion of the five most abundant genera of algae.
Ammonia-nitrogen, nitrate-nitrogen, nitrite-nitrogen, dissolved phos-
phorus, total phosphorus, and alkalinity were measured at NERC-Las Vegas
with a Technicon Autoanalyzer II according to the general methodology
described in Working Paper Number 1 (NES, 1974).
28
-------
The procedures used in the algal assay test, a methodology for the
determination of a water sample's productivity potential and limiting
nutrients, were those outlined in EPA's National Eutrophication
Research Program's publication entitled "Algal Assay Procedure Bottle
Test" (U.S. EPA, 1971). The analyses were conducted at PNERL, NERC-
Corvallis.
29
-------
SECTION IV
LAKE CLASSIFICATION
There is far from universal agreement as to what constitutes a lake. Veatch
and Humphrys (1966) suggested that to give the word "lake" a precise, limited
meaning would probably be an exercise in futility because the word has been
in use for a long time and been given a diversity of applications. The
word is used as a synonym for pond, reservoir, and sea. It has been applied
to bodies of fresh water and saline water; to standing water and widenings
in rivers; to bodies of water measuring less than a hectare and to those
gauged in hundreds of thousands of hectares; to naturally occurring Water
bodies and man-made reservoirs; to water-filled or partially filled basins;
and to basins void of water. "Lake" is generally more prestigious than
other common names (e_.g_., pond) and is preferred by promoters of water-
based tourist and recreational businesses and commercial developers of
shoreline property (Veatch and Humphrys, 1966). Nevertheless, numerous
attempts have been made to define and delimit the members of lentic series
(j_.e_., lake, pond, marsh, and their intergrades).
Fore! defined a lake as a body of standing water occupying a basin and
lacking continuity with the sea; he defined a pond as a lake of sliaht
depth (Welch, 1952). Welch (1952) defined a lake as a "...body of standing
water completely isolated from the sea and having an area of open, rela-
tively deep water sufficiently large to produce somewhere on its periphery
a barren, wave-swept shore." He employed the term "pond" "...for that class
of very small, very shallow bodies of standing water in which quiet water
and extensive occupancy by higher aquatic plants are common characteris-
tics" and suggested that all larger bodies of standing water be referred
to as lakes. Zumberge (1952) defined a lake as an inland basin filled
with water. Harding (1942) described lakes as "bodies of water filling
depressions in the earth's surface."
LAKES AS NATURAL RESOURCES
The exact number of lakes in the United States is unknown. Welch (1952)
estimated that there are at least 40,000 lakes in North America with a
combined surface area of no less than 225,000 square miles (583,000 square
kilometers). Hasler and Ingersoll (1968) reported a figure of 100,000
lakes in America (i_.e_., United States).
Estimates of lake numbers from different sources or compiled at different
times may exhibit wide disparity. The difference in estimates can be
related to the lack of agreement as to what constitutes a lake, incomplete-
ness of the inventories, and interpretatipnal differences of maps and
aerial photographs (Veatch and Humphrys, 1966). For example, Minnesota,
"The Land of 10,000 Lakes," completed an inventory of its surface waters
in 1968. The comprehensive final report initially defines a lake as
"... an enclosed basin, filled or partially filled with water..." and then
extends the term to include "...all natural enclosed depressions, 10 acres
30
-------
or more in area, which have substantial banks capable of containing water
and which are discernable on aerial photographs." Minnesota has, according
to the report, 15,291 lake basins of which 3,257 are partly or completely
dry (Minnesota Department of Conservation, 1968).
Fresh-water lakes are really insignificant when compared to the earth's
total surface as they account for about 0.16 percent of it (Table 4).
Nace (1960) estimates that only about 0.009 percent of the earth's total
water supply is in the form of fresh-water lakes. In some regions lakes
give the impression of dominating the landscape. Yet, in Beltrami County
(Minnesota), a place renowned for its abundance of lakes, only about 18
percent of the area is under water (Minnesota Department of Conservation,
1968). Vilas County (Wisconsin), another governmental unit well endowed
with lakes, has about 15 percent of its area covered by lakes (Deevey, 1942)
However, the geographic importance of any element of the landscape is not
measured merely on the basis of its areal extent. The intrinsic properties
of lakes make them a natural resource whose importance is greater than is
suggested by area alone. Lakes are used as sources of municipal water,
irrigation water, and cooling water for thermal-electric plants. They
serve as transportation routes and as focal points for many types of
recreational activity. Many lakes, particularly those in a pristine
condition, are valued for their aesthetic qualities. They also provide
convenient locations for dumping the organic and inorganic wastes of
society.
An increasing number of lakes are viewed as merely obstacles in the "way
of progress" and are being subjected to drainage or serving as land-fill
sites to provide additional farmland or building sites. As the world's
human population increases, it is likely that attempts to rid the land-
scape of lakes will increase in scope. Although drainage and land-fill
schemes threaten some lakes, a problem of much greater magnitude is that
of cultural (anthroprogenic) eutrophication.
LAKE SUCCESSION AND EUTROPHICATION
Lake Succession
Lakes, although giving the impression of permanence when measured on the
scale of the human life span, are transitory features of the earth's
surface. All lakes, regardless of their origin, pass through the process
of ecological succession which ultimately results in a terrestrial environ-
ment. The ephemeral nature of lakes is a consequence of two fundamental
processes, the downcutting of the outlet and, more important, the deposi-
tion of allochthonous and autochthonous materials in the basin.
Most lakes commence the successional process as bodies possessing rela-
tively low concentrations of nutrients and, generally, low levels of
31
-------
TABLE 4. DISTRIBUTION OF THE WORLD'S ESTIMATED WATER SUPPLY3
Location
World
Land Area
Surface water on the continents
Polar icecaps and glaciers
Fresh-water lakes
Saline lakes and inland seas
Average in stream channels
Total surface water
Subsurface water on the continents
Root zone of the soil
Ground water above depth 805m
Ground water, depth of 805m to
4,024 m
Total subsurface water
World's oceans
Total water on land
Atmospheric moisture
Total, world supply of water
Surface Area
(km2 x 103)
510,228
148,924
17,871
855
699
19,425
129,499
129,499
361,303
Volume of
Water
(km3 x 103)
30,428
125
104
1
30,658
25
4,168
4,168
8,361
1,321,314
39,014
12
1,360,340
Percentage
of Total
Water
2.24
0.009
0.008
0.0001
2.26
0.0018
0.306
0.306
0.61
97.1
2.87
0.001
'Adapted from Nace (1960).
-------
productivity.* The importation and deposition of materials (e_.c[., sediment)
from the shoreline and the surrounding watershed gradually decrease the lake
depth. The addition of allochthonous materials normally enriches the water
and thereby stimulates the production of organic materials. The autochtho-
nous materials increase the sedimentation rate thus accelerating succession.
Marked floral and faunal changes occur. Algal blooms become more common
along with submergent and eventually emergent aquatic macrophytes. Desir-
able game fish may be replaced by less desirable species, the so-called
"rough fish". A lake will eventually become a marsh or swamp which, in turn,
terminates as dry land.
Lindeman (1942) stressed the productivity aspects in relation to lake
succession. Figure 13 represents the probable successional productivity
relationships for a hypothetical hydrosere developing from a moderately
deep lake located in a fertile humid continental region. Productivity
is initially low, a consequence of low nutrient levels, but increases
rapidly as nutrients become more available. The length of time required
for completion of the successional process is a function of several factors
including lake basin morphology, climate, and the rate of influx and
nutrient value of allochthonous materials. It is readily apparent that
allochthonous nutrients can drastically increase lake productivity and
thereby shorten the life span of a lake.
The Concept of Eutrophication
The word eutrophication, until recently foreign to the vocabulary of the
general public, has appeared in the popular and scientific literature at
a rapidly increasing rate over the past decade. The term is often used
to denote the process whereby a pristine water body (e_.£., lake) is trans-
formed into one characterized by dense algal scums, obnoxious odors, and
thick beds of aquatic macrophytes. However, the word is applied differ-
ently, according to the respective interests of its users.
Weber (1907) used the German adjectival form of eutrophication (nahrstoff-
reichere - eutrophe) to describe the high concentration of elements
requisite for initiating the floral sequence in German peat bogs (Hutch-
inson, 1973). The leaching of nutrients from the developing bog resulted
in a condition of "mitterlreiche" (mesotrophe) and eventually "nahrstoff-
earme" (oligotrophe). Naumann (1919) applied the words oligotrophic
(under-fed), mesotrophic, and eutrophic (well-fed) to describe the nutrient
levels (calcium, phosphorus, combined nitrogen) of water contained in
springs, streams, lakes, and bogs (Hutchinson, 1973). Naumann (1931)
defined eutrophication as the increase of nutritive substances, especially
phosphorus and nitrogen, in a lake. Hasler (1947) broadly interpreted
eutrophication as the "Enrichment of water, be it intentional or uninten-
tional..." Fruh, et al_. (1966) defined the word as the "enhancement of
nutrients in natural water..." Edmondson (1974) suggested that many
limnologists seem to use the term to describe "an increase in the rate of
nutrient input..."
*Edmondson (1974) suggested that the idea, that all lakes are born oligo-
trophic and gradually become eutrophic as they age, is an old misconception.
33
-------
oo
EUTROPHY
VEGETATION
BOG FOREST
OLIGOTROPHY
MAT
SENESCENCE
TIME
Figure 13. Hypothetical productivity growth-curve of a hydrosere.
Adapted from Lindeman (1942).
-------
Hasler and Ingersoll (1968) suggested that eutrophication is the "process
of enrichment and aging* undergone by bodies of fresh water". Vollenweider
(1968) summarized the eutrophication of waters as meaning "... their
enrichment in nutrients and the ensuing progressive deterioration* of
their quality, especially lakes, due to the luxuriant growth of plants
with its repercussions on the overall metabolism of the waters affected..."
A search of the literature on eutrophication indicates that the meaning
of the term, originally limited to the concept of changing nutrient levels,
has been gradually expanded to include the consequences of nutrient enrich-
ment"1".
Eutrophication occurs both naturally and as a result of man's activities
(cultural or anthropogenic eutrophication). Many of man's practices
relating to the disposition of municipal sewage and industrial wastes
and land use impose relatively large nutrient loadings on many lakes and
rivers. In many cases, the enrichment results in algal blooms and other
symptoms of eutrophication. The consequences of man-induced eutrophication
often make the water body less attractive to potential users. More
importantly, at least when a long-range viewpoint is adopted, eutrophi-
cation accelerates lake succession, thus shortening the time period before
a lake loses its identity.
Trophic State and Trophic Indicators
Limnologists and other scientists concerned with lakes have used the
term "trophic state" (j_.e_., degree of eutrophy) to describe two different
lake characteristics, nutrient status and productivity. Thus, trophic
state is a hybrid concept as suggested by Margalef (1958). Several
different physical, biological, and chemical attributes are required to
adequately describe a lake's trophic state, making the concept multi-
dimensional (Brezonik and Shannon, 1971) and precluding its determination
through direct measurement. However, it is possible to quantify trophic
state through the use of trophic state indicators (indices) in conjunction
with appropriate data reduction techniques.
There are numerous indicators of trophic state, each with its merits and
shortcomings. Reviews have been written on the subject by Fruh, et_ al.
(1966), Stewart and Rohlich (1967), Vollenweider (1968), and Hooper TT969).
A list of some common indicators or indices are found in Table 5. A
diversity of opinion exists regarding the number and kinds of indicators
which should be considered in the classification of lakes. Use of a
* Emphasis added
+ The historical aspects and semantical problems associated with the word
"eutrophication" and its companion words (oligotrophtc, mesotrophic,
eutrophic) are found in Weber (1907), Naumann (1919, 1931), Thienemann
(1918), Rodhe (1969), Hutchinson (1967, 1973), Beeton and Edmondson (1972),
and Edmondson (1974).
35
-------
TABLE 5. TROPHIC INDICATORS AND THEIR RESPONSE TO INCREASED EUTROPHICATION9
Physical
Chemical
Biological
Transparency (d)
(Secchi disc reading )
Morphometry (d)
OJ
Nutrient concentrations (i)
(e...9_.» at spring maximum)
Chlorophyll a^ (i)
Conductivity (i)
Dissolved solids (i)
Hypolimnetic oxygen
deficit (i)
Epilimnetic oxygen
supersaturation (i)
Sediment type
Algal Bloom frequency (i)
Algal species diversity (d)
Littoral vegetation (i)
Zooplankton (i)
Fish (i)
Bottom fauna (i)
Bottom fauna diversity (d)
Primary production (i)
An (i) after an indicator signifies the value increases with eutrophication; a (d) signifies the value
decreases with eutrophi cation. The biological indicators all have associated qualitative changes (j_..e_.
species changes occur as well as quantitative (biomass) changes as eutrophication proceeds). Adapted
from Brezonik (1969).
-------
single indicator Kas the virtue of simplicity but may produce misleading
rankings or groupings because lakes are normally too complex to be
adequately gauged on such a simplified basis. On the other hand, the
use of a large number of indicators may result in a problem of character
redundancy.
MULTIVARIATE CLASSIFICATION OF LAKES*
A multiplicity of classificatory schemes has been devised to group and
rank lakes. Examples of some approaches to lake typology are found in:
Lueschow, et al_. (1970), Rawson (1956, 1960), Margalef (1958), Hansen
(1962), Jarnefelt (1958), Larkin and Northcote (1958), Moyle (1945,
1946), Pennak (1958), Round (1958), Whipple (1898, Winner (1972), ^afar
(1959), Beeton (1965), Donaldson (1969), and Gerd (1957). Hutchinson
(1957, 1967) has reviewed many of the attempts to arrange lakes into
orderly systems.
Lacustrine trophic state is a multi-dimensional concept and is, by its
very nature, amenable to analysis by multivariate statistical techniques
(e..£., cluster analysis, principal component analysis). Multivariate
techniques minimize the personal bias often present when data are examined
for groups, and rankings are developed. They are of particular value in
situations where large numbers of objects or parameters are to be
classified.
Numerical taxonomists and quantitative ecologists have been acutely
aware of the benefits which can be derived from multivariate techniques
and have been very active in promoting their use. Yet, a search of the
literature has yielded few publications in which the techniques have been
applied to lakes (£.£., Shannon, 1970; Brezonik and Shannon, 1971; Shannon
and Brezonik, 1972a, 1972b; Sheldon, 1972). Shannon and Brezonik
have devoted their efforts toward the classification and evaluation of
55 lakes in north central Florida. Sheldon reviewed the concept and
functions of classification, introduced multivariate techniques of
potential value in handling and synthesizing lake information, and applied
the techniques to several lake populations.
Lake scientists have been slow to apply multivariate techniques to the
problems of lake classification. This is probably because of a lack of
familiarity with the techniques, the unavailability of large digital
computers and/or the necessary software, and a shortage of comparable
data from large numbers of lakes .
The balance of this section is devoted to the application of two multi-
variate techniques (cluster analysis, principal component analysis) to a
* The term classification is often used in the restricted sense of placing
entities into distinct groups, thereby exluding arrangements showing no
distinct divisions (§_.£., ordination). The term is used here in the
broader context suggested by Sneath and Sokal (1973).
37
-------
group of 100 lakes (Table 3) using selected elements of ground truth
collected by NES. The resulting multfvariate trophic state index will
be used in Section VI to assist in the assessment of LANDSAT-1 MSS as a
tool in estimating the trophic status of lakes.
Cluster Analysis
Cluster analysis is a collective term encompassing a broad spectrum of <
techniques for delineating natural groups ("clusters") of objects or
attributes in hyperspace. A multitude of contributions have been published
on the subject '(.e.g., Ward, 1963; Lance and Williams, 1967, 1968; Gower,
1967; Padron, 1969). Three comprehensive publications, written by Sokal
and Sneath (1963), Anderberg (1973), and Sneath and Sokal (1973) serve as
excellent sources of information. An expose" of the various clustering
techniques available to researchers is beyond the scope of this investi-
gation, and the reader is referred to the above sources.
Clustering on objects is termed Q-technique as opposed to R-technique
which leads to classifications of attributes or characters (Sneath and
Sokal, 1973). Williams and Dale (1965) suggested that, when the relation-
ships are represented in hyperspace, the kind of space that is operated
on should be called A-space and I-space, not R-space or Q-space. A-space
(attribute space) has p dimensions, one for each attribute (character), in
which there are n points, each representing an object. I-space (individual
space) has n dimensions, one for each object, in which there are p points,
each representing an attribute. This study will utilize the Q-technique
operating in A-space.
Objects and Attributes
One hundred lakes, sampled by NES in 1972, were selected for analysis. Each
lake was assigned a serial number (1-100) which is unique to this report. A
careful examination of the physical, chemical, and biological parameters
measured by NES resulted in the selection of six indicators for incorpora-
tion into both the cluster analysis and principal component analysis (PCA)
ordination. A seventh indicator, an abbreviated form of the Pearsall
cation ratio inverse was originally included, but was eliminated because
it did not appear to contribute significantly to the classification scheme*
The indicators (I.e.., lake attributes) are: conductivity (COND, ymhos
cm"1), chlorophyll a^ (CHLA, yg T1)* total phosphorus (TPHOS, mg
r1), total organic nitrogen (TON, mg 1-1)> algal assay yield (AAY, dry-
wt in mg) and Secchi disc transparency (SECCHI, m). The inverse of Secchi
disc transparency (ISEC, m'1) was employed so that all of the indicator
values would increase as trophic status increases.
The six indicators were selected because they are quantitative, considered
to be important measures of trophic state, and satisfy Hooper's (1969)
* Abbreviated form of cation inverse is: ICAT = sodium / (calcium +
magnesium).
38
-------
criteria for trophic indices. Annual mean values for COND, CHLA, TPHOS,
and ISEC were used in the analyses; AAY and TON measurements were limited
to the fall overturn sample, precluding the use of an annual mean. The
descriptive statistics of the six indicators are found in Table 6. A
serious lack of normality in the data necessitated a transformation prior
to clustering. Natural logarithms (LN) were found to be adequate for this
transformation.
Cluster Method
A complete linkage algorithm (McKeon, 1967) was used to examine the
lakes for natural groupings. The method is also known as the furtherest
neighbor hierarchical strategy (Lance and Williams, 1967) and the maximum
method (Johnson, 1967). The algorithm is characterized as agglomerative,
nonoverlapping, and hierarchical, the latter property permitting the out-
put to be modelled in the form of dendrograms. The program was run on an
International Business Machines (IBM) Model 370-155E digital computer at
Optimum Systems Incorporated (OSI), Bethesda, Maryland.
The LN-transformed trophic indicator data were entered as an N x p matrix
where N is the number of observations or objects (I.e.., lakes) and p is
the number of dimensions or attributes (i.e.., indicators). The data points
in each column (indicator) of the data matrix were standardized by sub-
tracting the column mean from each point and then dividing by the column's
standard deviation. This was accomplished using the McKeon program option
number three. Standardization was necessary because the data were measured
in different units.
Euclidian distance, an extension of the Pythagorean theorem to points in
hyperspace, was selected as the similarity coefficient, largely because
of its intuitive appeal. The formula for the Euclidian distance, Ajk,
between two objects (e_.£., lake "j" and lake "k") -js
A
jk
The distances between all possible (N(N-l)/2) pairs of objects are computed
and stored in an N x N symmetrical matrix (S-matrix). The McKeon program
uses the squared Euclidian distance, A2, but the same clusters could be
obtained by using Euclidian distance or any monotonic function of that
distance.
The clustering procedure carries out successive iterations on the S-matrix
which has the maximum squared distance within cluster in the diagonal and
the maximum squared distance between clusters in the off-diagonals. Initially
the S-matrix is N x N with zeros in the diagonals (Figure 14). At each
39
-------
TABLE 6. DESCRIPTIVE STATISTICS OF 100 LAKES9
-p.
o
Statistic
Mean
Median
Maximum value
Minimum value
Range
Standard deviation
Skewness
CHLA
(yg i~ )
22.3
11.8
381.2
1.1
380.1
42.7
6.4
ISEC
0.85
0.54
5.37
0.13
5.24
0.88
2.86
COND
Cijmhos cm'1)
352
353
808
50
758
155.9
0.3
TPHOS
(mg I'1)
0.136
0.048
1.893
0.005
1.888
0.265
4.264
TON
(mg I-1)
0.80
0.67
3.35
0.08
3.27
0.56
2.20
AAY
(mg dry-wt)
7.3
2.7
61.3
0.1
61.2
10.6
2.4
a The values above represent the statistics for the NES lakes which are .incorporated into
the cluster and principal component analyses.
-------
0,
2
3
4
oo
LJ
N
LAKES
234
N
0,
A231 A23:
A241 A24:
THE UPPER TRIANGULAR FORM OF THE
MATRIX IS A REFLECTION OF THE
LOWER TRIANGULAR FORM
FOR EACH PAIR
OF LAKES
Q
Figure 14. N x N S-matrix. Initially the matrix has only zeros on
the diagonal. The off-diagonal elements represent the
squared Euclidian distances between each pair of lakes.
There are 4,950 distances possible for 100 lakes.
Appendix E contains the matrix of A2 used to examine
the NES-sampled lakes for clusters.
41
-------
iteration, those two clusters are combined, which, taken together, form
the most compact cluster. The measure of compactness is the maximum
distance between any two points (j_..e., lakes) within the cluster. The
next pair to be combined are identified by finding the smallest squared
distance between points in the off-diagonals.
For example, if N=100, the program will start out with 100 clusters and
successively meld the clusters, two per iteration, eventually terminating
with one cluster containing 100 objects.
Results and Discussion
The results of the cluster analysis are depicted as a dendrogram (Figure
15). The abscissa is scaled in Euclidian distance with the points of
junction between stems implying that the maximum within-cluster distance
is the value on the abscissa. Generally, the ordinate of a dendrogram
has no special significance. The order of the clusters can be changed by
rotating the stems, thereby producing a multitude of apparently^ different
dendrograms. Interpretational problems arise when it is necessary to
compare dendrograms developed from different data or algorithms.
Some investigators attempt to "standardize" the ordinate-induced appear-
ance of the dendrogram by rotating the stems to keep the code or serial
numbers as ordered as possible, especially if the numbers present linear
arrangements of generally established taxonomic groups (Sneath and Sokal,
1973). The author has attempted to order the axis by rotating the stems
to reflect the PCA ordination results found in the next section of this
chapter. Consequently, the trophic status of the 100 lakes in the dendro-
gram generally increases along the ordinate in the "downward" direction.
It has become an accepted practice in lake studies to use the terms oligo-
trophic, mesotrophic, and eutrophic in reference to the trophic status
of lakes. The terms, although well established in the literature, are
used freely and it is difficult to quantitatively determine what is meant
by them (Beeton, 1965). It may be argued that the terms serve to stereo-
type lakes and unduly restrict, to three categories, what may be members
of a trophic continuum. However, this terminology is likely to continue
in vogue because, at a very minimum, it gives some indication of trophic
state.
The question arises regarding the number of "good" natural clusters which
have been depicted by the clustering algorithm. McKeon (1967) asserted
that a sudden increase in the maximum within cluster distance suggests
that the previous stage may be a good stopping point. The first "sudden"
increase occurs between A3.70 and A4.35 indicating that there may be seven
relatively coherent clusters (A, B, C, D, E, F, 6). There is some diffi-
culty involved in reconciling the seven clusters with the three classic
states.
42
-------
A
A
Figure 15. Dendrogram of 100 lakes sampled by the National
Eutrophication Survey during 1972. The dendrogram is
based on a complete linkage algorithm using generalized
Euclidian distance as the measure of similarity.
43
-------
The NES, cognizant of the advantages and limitations of naming lakes
according to a three-class trophic scheme, has applied the trophic names
to these lakes using data collected in 1972 and information acquired from
various sources, including reports and knowledgeable individuals. Using
the NES assessments as a guide, Clusters A and B may be characterized as
containing a mixture of oligotrophic and mesotrophic lakes. Cluster C
consists of eutrophic lakes; Cluster D is comprised of both mesotrophic
and eutrophic lakes. Eutrophic lakes made up Clusters E and F- Cluster
G consists of lakes which are very eutrophic (hypereutrophic).
The results of the cluster analysis appear distressing in light of the
three-class concept of trophic state. A more careful selection of candi-
date lakes could have resulted in the "discovery" of three groups matching
the classic trophic states. Nevertheless, the clustering approach is of
value in showing relationships among lakes. It gives the investigator
another way of perceiving his study lakes, and it is hoped will enable
him to see more clearly the relationships between large numbers of lakes.
Principal Components Ordination
Hierarchical methods are a rather heavy-handed approach to the problem
of reducing the dimensionality of multidimensional systems (Sheldon, 1972).
They have the inherent capability to yield some form of clusters regardless
of the structure of the data constellation, even if the entities to be
analyzed are randomly distributed (Sneath and Sokal, 1973). Another
approach that merits consideration is ordination. Ordination is the place-
ment of N entities in A-space varying in dimensionality from 1 to p or
N-l, whichever is less. Principal components analysis, one ordination
technique, will be used to examine the lakes in A-space for natural clusters
and to derive a multivariate trophic state index.
Principal components analysis may be used to reduce the dimensionality
of a multivariate system by representing the original attributes as functions
of the attributes. The main object is to summarize most of the variance
in the system with a lesser number of "artificial" variates (j_..e., prin-
cipal components).
The computation of principal components can be undertaken using either
a covariance matrix (S) or a p x p matrix of Pearson product-moment corre-
lation coefficients (R). Use of the R-matrix is indicated when the variates
are measured in different units (.§_.£., grams and meters). Computation of
the R-matrix principal components involves the extraction of its eigen-
values (characteristic or latent roots) and eigenvectors (characteristic
or latent vectors). The eigenvalues are a set of r nonzero, positive
scalar quantities. The sum of the R-matrix eigenvalues is the matrix
trace and is equal to the number of dimensions in the original system
(j_.£., the number of variates, p). The rank of the matrix is r and is
equal to p.
44
-------
Normalized eigenvectors give the A-space coordinates of an orthogonal
set of axes known as the principal axes. The normalized eigenvectors
are commonly designated as principal components.
The first principal component of the observations of the p-variates
X-|, ..., Xp is the linear compound
Y!=ail XT + ... + ap1 Xp
whose coefficients (a-ji) are the elements of the eigenvector associated
with the largest eigenvalue of the R-matrix (Morrison, 1967). The variance
of the first principal component is associated with the eigenvalue. The
jth principal comoonent of the p-variate system is the linear compound
Yj^ijXl + ... + apj Xp
whose coefficients are the elements of the eigenvector associated with
the jth largest eigenvalue extracted from the R-matrix. The jth eigenvalue
is a measure of the variance of the jth principal component.
The proportion of the total sample variance in the cloud of dimensionless
standard scores attributable to any component is found by dividing its
eigenvalue by p. The first principal component has the innate property
of explaining the greatest proportion of the sample variance with each
successive component explaining progressively smaller amounts of the total
sample variance. Frequently, a consequence of the decreasing order of
variance, k < r dimensions will adequately summarize the variability
of the original variates X-j, ..., Xp. The first three components generally
account for most of the variation permitting the ordination of the subjects
in 1-D, 2-D, and 3-D space. All of the dispersion in the data can be
accounted for by using r dimensions, but this negates the analysis objective,
the reduction of dimensionality or as Seal (1964) stated the "... parsimo-
nious summarization of a mass of observations."
The principal components of N p-variate observations may be defined geomet-
rically (Morrison, 1967) as "... the new variates specified by the axes of
a rigid rotation of the original response coordinate system into an orien-
tation corresponding to the directions of maximum variance in the sample
scatter configuration." The normalized eigenvectors give the directions
of the new orthogonal axes and the eigenvalues determine the lengths
(j_.e_., variance) of their respective axes. The coordinate system is
expressed in standard units (zero mean, unit variances) when the compon-
ents are extracted from the R-matrix. Figure 16 is a hypothetical bivar-
iate example of the geometric meaning of principal components.
Detailed descriptions of the theoretical and computational aspects of
principal components are found in Retelling (1933a, 1933b, 1936), Anderson
(1958), and Morrison (1967).
45
-------
FIRST PRINCIPAL AXIS
CORRESPONDING TO THE
FIRST PRI N Cl PAL
COMPONENT I
Y, oX + bX.
SECOND PRINCIPAL AXIS
CORRESPONDING TO THE
SECOND COMPONENT S
Y, - c X, d X,
CHLOROPHYLL a (X,)
Figure 16. Geometrical interpretation of the principal components for a
hypothetical bivariate system. Principal components may be
interpreted geometrically as the variates corresponding to
the orthogonal principal axes of observation scatter in A-
space. The elements of the first normalized eigenvector
(l.e_., coefficients of the first principal component) define
the axis which passes through the direction of maximum vari-
ance in the scatter of observations. The associated eigen-
value corresponds to the length of the first principal axis
and estimates the dispersion along it. The second principal
component corresponds to the second principal axis, the length
of which represents the maximum variance in that direction.
In our example the first component accounts for most of the
dispersion in the data swarm and the original 2-dimensional
system could be summarized in one dimension with little loss
of information. The new variate value (PCI) for each lake is
obtained by evaluating the first component
Y! = aX
1
bX2
The PCI for each lake in 1-D A-space is its coordinate on the
first component axis and is shown diagrammatical!,/ by project-
ing each observation to the principal axis. (Modified from
Brezonik and Shannon, 1971).
46
-------
Methodology
The principal components analysis was accomplished using the same 100
lakes and LN-transformed trophic state indicator data used in the hierar
chical cluster analysis (LNCHLA, LNISEC, LNCOND, LNTPHOS, LNTON, LNAAY).
The data matrix was further standardized (zero mean, unit variance) by
attributes using the relationship
Z-; -; =
where zn- ^
(i.e.,
Xj and
is
the standardized values for attribute
j; XT n- is the LN-transformed value of
i of observation
lake) j; X-jj is the LN-transformed value of observation j; and
s.j are the mean and standard deviation of attribute j, respectively.
The eigenvectors and eigenvalues were extracted from a p x p correlation
matrix (Table 7). All of the computational aspects were executed on a
Control Data Corporation digital computer (CDC 3300) at Oregon State
University using the Statistical Interactive Programming System (SIPS).
A detailed explanation of SIPS, and its operation is found in Guthrie,
Avery, and Avery (1973).
TABLE 7. R-MODE CORRELATION MATRIX OF SIX TROPHIC STATE INDICATORS.
The coefficients were determined using LN-transformed
data for the 100 NES-sampled lakes.
LNCHLA LNISEC LNCOND LNTPHOS
LNTON LNAAY
LNCHLA
LNISEC
LNCOND
LNTPHOS
LNTON
LNAAY
1.000 0.886 0.397
1.000 0.285
1.000
0.801
0.777
0.397
1.000
0.684
0.660
0.378
0.683
1.000
0.686
0.597
0.327
0.810
0.576
1 .000
As was asserted earlier, principal components analysis can be used to
advantage because often k <_ 3 dimensions will explain most of the variance
in the hyper-dimensional cloud of data points. The resulting ordinations
can be expressed as 1-D, 2-D, and 3-D models. The ordinations are usually
expressed as a sequential numerical listing (i_.e_., 1-D model), a scatter
diagram (i_.e_., a 2-D model), or as a set of three 2-D scatter diagrams
which, when examined carefully, may give some indication of the scatter
of observations in an A-space of three dimensions. However, unless the
47
-------
pattern of diversity is simple, it is rather difficult for most individuals
to visualize the pattern in 3-D space using a series of 2-D projections.
Fraser and Kovats (1966) and Rohlf (1968) have advocated the use of stereo-
graphic projections for ordination in 3-space and have furnished the equa-
tions necessary for the development of stereoscopic models (i.e., stereo-
grams). Some examples are found in Schnell (1970), Moss (19677, and Sneath
and Sokal (1973).
This investigator has employed the techniques in Rohlf (1968) to construct
3-dimensional ordination models by plotting the lakes in the 3-D space
produced by the first three principal components. The models can be
examined visually for the presence of clusters. Unlike the clustering
approach, no assumptions are made that the lakes must congregate into
a series of clusters.
the model, as used here, consists of one member of a stereo pair; the
other member is easily produced if the complexity of the data necessi-
tates the advantages inherent in viewing the objects in stereo. The lake
coordinates, determined by evaluating the three components for each
lake, are standardized to make the scale for the longest axis (X) run
from 0.0 to 1.0. The other axes are scaled-down proportionally. The
component with the smallest range is assigned to the vertical (Z) axis.
The models were plotted on a Calcomp (IBM-1724) 30 inch incremental
drum plotter driven by the Oregon State University CDC 3300 computer
using subroutines found in GRAFPAC, a plotting routine package developed
by Rohlf (1968). The same modelling approach is used in Section VII to
depict LANDSAT-1 MSS color ratio relationships.
Results and Discussion
The normalized eigenvectors and eigenvalues are found in Table 8. Although
the principal component analysis is of value in reducing the dimensionality
of a multivariate system, it is sometimes difficult to interpret the new
variates in terms of subject matter identities. Some indication of a
principal component's meaning may be ascertained by an examination of the
algebraic sign and magnitude of its coefficients.
The coefficients of the first component (Table 8), excluding the coeffic-
ient for LNCOND, are nearly equal in magnitude suggesting that it represents
a general measure of trophic state, accounting for approximately 68 percent
of the variation in the data. Correlations between the new variate and
the LN-transformed trophic indicators are found in Table 9.
The first principal component was evaluated for each of the 100 lakes.
The resultant values (PCI) are indicative of each lake's respective
position on a multivariate trophic scale (Table 10). The procedure
followed is essentially that of Brezonik and Shannon (1971), but the
48
-------
TABLE 8. NORMALIZED EIGENVECTORS AND EIGENVALUES3
Eigenvector
Number LNCHLA
1 0
2 -0
3 -0
4 -0
5 -0
6 -0
.457
.112
.222
.346
.436
.647
LNISEC
0.435
-0.249
-0.397
-0.395
0.060
0.657
LNCOND
0.251
0.952
-0.018
-0.143
-0.006
-0.099
LNTPHOS
0.458
-0.088
0.271
-0.040
0.797
-0.270
LNTON
0.403
0.010
-0.383
0.829
-0.056
0.008
LNAAY
G.408
-0.104
0.757
0.122
-0.410
0.258
Eigenvalue
4.081
0.810
0.477
0.387
0.152
0.092
Variance
(*)
68
13
7
6
2
1
100
.02
.50
.95
.45
.54
.54
Cumulative
Variance (%)
68
81
89
96
98
100
.02
.52
.47
.92
.46
.00
principal components analysis was performed using an R-matrix of correlation coefficients for
six trophic state indicators. The data represent 100 lakes sampled by NES during 1972.
-------
scale was not shifted into the positive domain by correcting each lake's
PCI with the PCI obtained from a hypothetical lake. The lake lying at
the negative end of the scale, Crystal Lake, is rated as having the lowest
trophic status of those studied; in other words, it has relatively high
water quality. Trophic state increases in the positive direction on the
scale with the lake lying at the positive extreme, Albert Lea Lake, exhib-
iting the highest trophic state of the 100 lakes studied. Some breaks or
gaps are evident on the scale, but only toward the two extremes. A very
discernible gap occurs between Beaver Dam Lake (53, position 95) and Mud
Lake (24, position 96); the Mud Lake - Albert Lea Lake group might be
termed as hypereutrophic.
TABLE 9. PRODUCT-MOMENT CORRELATION COEFFICIENTS OF THE TROPHIC STATE
INDICATORS AND THE PRINCIPAL COMPONENTS.
Principal Component
3 4
LNCHLA
LNISEC
LNCOND
LNTPHOS
LNTON
LNAAY
0.92
0.88
0.51
0.92
0.82
0.83
-0.10
-0.22
0.86
-0.08
0.01
-0.09
-0.15
-0.27
-0.01
0.19
-0.26
0.52
-0.22
-0.25
-0.09
-0.03
0.52
0.08
-0.17
0.02
-0.20
0.31
-0.02
-0.16
-0.20
0.20
0.01
-0.08
0.00
0.08
The magnitude,of the first component's coordinates suggests that a less
elegant approach toward an ordination of the lakes might be undertaken
with similar results. The approach, using the same LN-transformed stan-
dardized indicator values, involves the summation of a lake's indicator
values and then dividing by p, the number of indicators, In other words
X^ is the jth
lake's score on
MCRi =
J
where MCRj is the Mean Composite Rank for the jth lake,
lake's score on the first trophic indicator, Xjp is the
the pth indicator and p is the number of dimensions or indicators. The
results of this ranking method are found in Table 10 along with the
principal components ordination. The two methods of ordinating the lakes
are in close agreement. However, the principal component approach has the
advantage of permitting the development of 2-dimensional and 3-dimensional
ordinations which explain a very high percentage of the variance.
50
-------
TABLE 10. PRINCIPAL COMPONENT ORDINATION AND MEAN COMPOSITE RANK ORDINATION OF 100 LAKES
CJl
Position
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Lake Name
Crystal
Schroon
Higgins
Canandaigua
Charlevoix
Trout
Seneca
Cayuga
Crystal
Owasco
Mi ddl e
Keuka
Round
Oconomowoc
Geneva
Green
Hough ton
Carlos
Lac La Belle
Leech
Conesus
White Bear
Birch
Pelican
Forest
Rock
Serial
Number
44
87
70
92
69
43
94
95
77
96
65
93
50
58
68
51
71
17
57
6
91
31
7
11
30
55
PCI
Value
-5.04
-4.59
-3.92
-3.63
-3.60
-3.24
-2.89
-2.74
-2.52
-2.47
-2.29
-2.14
-2.09
-1.82
-1.71
-1.67
-1.61
-1.55
-1.43
-1.43
-1.41
-1.41
-1.39
-1.27
-1.22
-1.21
Lake Name
Crystal
Schroon
Higgins
Canandaigua
Charlevoix
Trout
Seneca
Cayuga
Owasco
Crystal
Middle
Keuka
Round
Hough ton
Pelican
Oconomowoc
Geneva
Green
Long
Birch
Leech
Carlos
White Bear
Conesus
Lac La Belle
Forest
Serial
Number
44
87
70
92
69
43
94
95
96
77
65
93
50
71
11
58
68
51
41
7
6
17
31
91
57
30
MCR
Value
-2.17
-1.99
-1.56
-1.42
-1.41
-1.39
-1.03
-1.02
-0.98
-0.95
-0.85
-0.85
-0.79
-0.66
-0.65
-0.64
-0163
-0.62
-0.59
-0.58
-0.57
-0.57
-0.55
-0.54
-0.51
-0.49
-------
TABLE 10. PRINCIPAL COMPONENT ORDINATION AND MEAN COMPOSITE RANK ORDINATION OF 100 LAKES (continued)
en
ro
Position
27
28
29
30
31
32
33
34
35
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
Lake Name
Green
Long
Browns
Le Homme Dieu
Cassadaga
Trout
Shawano
Cass
Black
Darling
Pine
Chautauqua
Okauchee
Shagawa
Saratoga
White
Rabbit
Bemi d j i
Cranberry
Whitewater
Minnewaska
Andrusia
Yellow
Pere Marquette
St. Croix
Serial
Number
21
41
64
18
89
8
46
5
88
16
60
90
59
12
100
73
14
2
15
10
19
3
39
72
32
PCI
Value
-1.10
-1.07
-1.07
-1.06
-1.04
-1.01
-0.94
-0.90
-0.77
-0.73
-0.71
-0.66
-0.62
-0.58
-0.57
-0.47
-0.46
-0.44
-0.40
-0.38
-0.32
-0.24
-0.23
-0.21
-0.17
Lake Name
Cassadaga
Rock
Black
Le Homme Dieu
Green
Shagawa
Shawano
Trout
Cass
Gull
Chautauqua
Cranberry
Pine
Saratoga
Darling
Whitewater
Elk
Okauchee
Rabbit
Yellow
Bemidji
St. Croix
White
Wapogasset
Andrusia
Serial
Number
89
55
88
18
21
12
46
8
5
13
90
15
60
100
16
10
42
59
14
39
2
32
73
40
3
MCR
Value
-0.50
-0.44
-0.42
-0.41
-0.41
-0.40
-0.40
-0.38
-0.36
-0.34
-0.34
-0.30
-0.26
-0.25
-0.24
-0.24
-0.22
-0.18
-0.17
-0.17
-0.16
-0.14
-0.13
-0.12
-0.11
-------
TABLE 10. PRINCIPAL COMPONENT ORDINATION AND MEAN COMPOSITE RANK ORDINATION OF 100 LAKES (continued)
in
GO
Position
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
Lake Name
Wapogasset
Elk
Chemung
Blackduck
Clearwater
Thompson
Wolf
Otter
Muskegon
Union
Ta inter
Thornapple
Pewaukee
Swan
Minnetonka
Maskenode
Nest
Cross
Strawberry
Round
Long
Como
Butte des Morts
Nagawicka
Ford
Madison
Serial
Number
40
42
81
1
23
82
4
98
74
84
45
79
62
52
29
9
20
97
80
99
85
67
48
61
83
35
PCI
Value
-0.16
-0.08
-0.06
0.01
0.01
0.03
0.11
0.13
0.18
0.43
0.44
0.58
0.59
0.68
0.73
0.74
0.77
0.86
0.99
1.02
1.14
1.15
1.27
1.27
1.36
1.36
Lake Name
Pere Marquette
Minnewaska
Blackduck
Chemung
Wolf
Otter
Clearwater
Thompson
Muskegon
Ta inter
Union
Mashkenode
Pewaukee
Swan
Minnetonka
Thornapple
Nest
Round
Cross
Strawberry
Como
Butte des Morts
Long
Madison
Nagawicka
Sakatah
Serial
Number
72
19
1
81
4
98
23
82
74
45
84
9
62
52
29
79
20
99
97
80
67
48
85
35
61
36
MCR
Value
-0.04
-0.03
-0.02
0.02
0.02
0.03
0.03
0.08
0.09
0.10
0.22
0.27
0.29
0.30
0.30
0.30
0.32
0.37
0.41
0.45
0.46
0.48
0.50
0.53
0.56
0.57
-------
TABLE 10. PRINCIPAL COMPONENT ORDINATION AND MEAN COMPOSITE RANK ORDINATION OF 100 LAKES (continued)
Ol
-pa
Position
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
Lake Name
Sakatah
Jordan
Kegonsa
Cokato
Poygan
Randall
Delavan
Pepin
Mona
Buffalo
Spring
Winnebago
Koshkonong
Fremont
Tichigan
Bear
Beaver Dam
Mud (at Maple Lake)
Wagonga
Carrigan
Silver
Albert Lea
Serial
Number
36
78
54
25
47
86
66
34
76
26
33
49
56
75
63
37
53
24
22
27
28
38
PCI
Value
1.38
1.41
1.48
1.61
1.68
1.92
2.03
2.10
2.17
2.31
2.33
2.36
2.45
3.06
3.11
3.15
3.29
4.32
4.40
4.40
4.79
5.90
Lake Name
Jordan
Kegonsa
Ford
Poygan
Cokato
Randall
Delavan
Pepin
Mona
Winnebago
Buffalo
Spring
Koshkonong
Bear
Fremont
Tichigan
Beaver Dam
Mud (at Maple Lake)
Carrigan
Wagonga
Silver
Albert Lea
Serial
Number
78
54
83
47
25
86
66
34
76
49
26
33
56
37
75
63
53
24
27
22
28
38
MCR
Value
0.58
0.59
0.60
0.65
0.71
0.80
0.83
0.86
0.87
0.91
0.92
0.94
1.02
1.25
1.25
1.27
1.29
1.70
1.76
1.79.
1.89
2.34
-------
The second component (Table 8) explains about 14 percent of the variation
in the data. The LNCOND coordinate of the component is very large sugges-
ting that the second variate is largely a measure of conductivity. The
second component has a good correlation with LNCOND (Table 9).
The third component (Table 8) accounts for approximately 8 percent of the
variance. LNAAY is the trophic indicator which shows the strongest corre-
lation with the third component (Table 9), but is not readily interpretable.
Approximately 89 percent of the total sample variance can be attributed to
the first three components. Figure 17 depicts the 100 lakes ordinated in
3-D space defined by the first three components. Some small clusters are
apparent, but they are not well defined. The long axis is the first
principal component axis (I). The axis labelled II is the second component
and the vertical axis (III) is the third component.
The failure to discern well defined clusters may be partially a conse-
quence of the trophic state of the lakes incorporated into this analysis;
the NES lake population is heavily weighted toward lakes having water
quality problems.
Summary
One hundred lakes were subjected to two complementary multivariate analyses,
a complete linkage hierarchical cluster analysis and ordination using the
technique of principal components. Well-defined clusters or natural group-
ings were not found through either approach. This may be a consequence,
at least partially, of the "kinds" of lakes incorporated into the analyses.
The first principal component was evaluated for each lake and the resulting
value (PCI), its coordinate on the axis, used as a multivariate index of
the lake's trophic state. The PCl's are purported to represent an assess-
ment of lacustrine trophic state and will be used in Section VII to evaluate
LANDSAT-1 MSS color - lake relationships.
55
-------
en
cn
Figure 17. The three-dtmensional principal component ordination of 100 lakes sampled by the National
Eutrophication Survey during 1972.
-------
SECTION V
LANDSAT-1 DATA EXTRACTION TECHNIQUES AND PRODUCTS
LANDSAT-1 MSS data are available from EROS in the form of photographic
products and computer-compatible digital magnetic tapes (CCT's). The
data forms permit investigators two general approaches to the problem
of data extraction and utilization, the photographic approach and the
CCT approach. Each has its particular advantages and limitations.
MSS DATA EXTRACTION APPROACHES
Photographic Approach
The photographic products available from EROS include black and white
(b&w) negative and positive transparencies, and prints. False color
prints and transparencies are available for a very limited number of
scenes. All are either third or fourth generation photographic products.
The photographs may be examined visually, as was done in Section III,
for differences in gray tone, texture, shape, size, and pattern. Tech-
niques common to the interpretation of aerial photographs may be
employed to extract information pertinent to many different fields of
science.
Quantification of the information contained in the undodged photographic
products may be achieved through microdensitometry and photographic
techniques including false color enhancement and density slicing.
Manual and automatic microdensitometry can be used to digitize the
information contained on the photographic positive or negative transpar-
encies. This is easily accomplished with a microdensitometer by
directing a light beam of known intensity through a small segment
(e_.£., lake image or portion of a lake image) of the transparency and
measuring any changes in intensity on a numerical scale as percent
transmission or as optical density. The quantitative data may then be
used to examine correlations between MSS bands and lake indicators
relating to trophic state.
Automatic scanning systems permit density slicing, the separation of the
different densities on a transparency and their coding for later repro-
duction as b&w and color-enhanced products or symbolized listings.
Reliance upon sophisticated instrumentation for the determination of
areas of equal density (i.e.., equidensities) can be avoided by using
various photographic enhancement techniques. Nielsen (1972) summarizes
a simple process reported by Ranz and Schneider (1970) which utilizes
Agfacontour film for the production of equidensities.
57
-------
The photographic approach is attractive because it can be accomplished
with relatively simple equipment (unless a fully automated scanner is
used) and is inexpensive. The four transparencies required to give all
band (green, red, IR1, IR2) coverage of a LANDSAT-1 MSS scene cost
only ten percent of the listed (November 1975) EROS price for a compar-
able set of CCT's ($20.00/$200.00). Many investigators, lacking the
necessary computer and software for processing the CCT's, are able to
explore possible applications of LANDSAT-1 data in their areas of
specialization using the photographic approach.
However, the approach has several limitations which are particularly
serious in water resources studies and which merit mention. The
transparencies have a relatively small density range when compared to
the sensitivity range of the MSS; this results in a scale compression
when the MSS data are transformed into a film image on an electron
beam recorder (EBR). The range of energy returns from water bodies
is small and located at the lower end of the MSS intensity scale.
Compression of the overall scale increases the difficulty of discrimi-
nating differences in water quality.
As mentioned previously, the photographic materials are third and
fourth generation products (the first generation is the film image
produced by the EBR), and the errors common to photographic processing
are compounded. Chemical adjacency effects and the nonlinearity of the
density-log exposure (D-log E) curve increase the magnitude of the
problem. An additional element of uncertainty is introduced by the
microdensitometer.
This investigator is not trying to assert that the photographic approach
is totally lacking in merit, but rather that its limitations seriously
limit the investigator attempting a quantitative estimation of lake
parameters relating to trophic indicators and trophic state.
CCT Approach
The use of computer compatible tapes to study LANDSAT-1 MSS sensed
phenomena is not only more expensive, but it is also more restrictive
because there are relatively few research centers with the requisite
computer software and output devices to use the digital data to full
advantage. However, these limitations are overshadowed by the fact
that the tapes contain, in digital form, the actual data point values
recorded by the MSS. Investigators can avoid the numerous uncertainties
introduced when the original MSS data are coded by an electron beam
recorder (EBR) into photographic forms and then re-quantified through
microdensitometry.
This investigation was initiated using the photographic approach. A
manually operated microdensitometer was used to measure the optical
density (OD) of lake images. Differences in OD were detected, but
58
-------
several difficulties were encountered (e_.£., variation in the quality
of the transparencies, locating the lakes on the green band transparen-
cies), and the approach was abandoned in favor of the CCT approach.
The CCT approach permits the rapid determination of picture element
(pixel) counts, descriptive statistics (e_.cj_., means, standard deviations,
and histograms), density slicing, and affords the opportunity to
enhance optional photographic products through both linear and non-linear
contrast stretching.
The utilization of CCT's is not without its problems, most of which are
related to the quality of the information contained on the tapes.
Defects commonly contained in the MSS data recorded on the tapes include
missing lines and random bit-dropout for major portions of lines. The
defective tapes can be repaired using existing software, but at addi-
tional cost both in time and money. Another problem is the receipt of
CCT's not meeting the specification of the purchase order (e_.£., receipt
of 9-track instead of 7-track tapes; 556 bpi instead of 800 bpi tapes).
While this is a minor problem, it does cost the investigator valuable
time.
CCT DATA EXTRACTION TECHNIQUES
The LANDSAT-1 MSS CCT's used in this investigation were processed at the
Image Processing Laboratory (IPL), a support facility of the Jet Propul-
sion Laboratory (JPL), California Institute of Technology. In this
section, the author will briefly discuss the image processing system
and, using one lake as an example, outline the data extraction tech-
niques employed.
Image Processing System
The IPL system consists of an IBM 360/44 computer with five tape drives,
four disc drives, a quick-look Polaroid pictorial output device, and a
core-refreshed interactive display (Blackwell and Boland, 1974). A
video film converter (VFC) provides offline hardcopy pictorial input and
output capabilities. This device has a precision cathode ray tube (CRT),
and 70-mm and 35-mm cameras. In the pictorial output mode, the VFC
reads a digital magnetic tape, containing a digital image, and displays
that image on the CRT, exposing the film which is then developed and
printed, producing a hardcopy output of the digital image. In the
pictorial input mode, the VFC functions as a flying spot scanner pro-
ducing a digital output picture on magnetic tape.
The IPL IBM 360/44 operates under the control of a special software
system, VICAR (Video Information Communication and Retrieval). Presently,
a highly stylized Operating System (OS) is operational which permits
foreground-background, roll-in and roll-out capabilities. VICAR is
designed to allow very flexible manipulation of digital pictures
consisting of rectangular arrays of optical measurements. The system
59
-------
contains a library of more than two hundred operationally executable
image-processing programs. Each program operates on an input digital
picture stored on magnetic tape or disc. Processing line-by-line, it
generates an output picture on tape or disc, preserving the input data.
Additional information regarding the system and existing programs is
found in Frieden (1971), Anon (1973), and Blackwell and Billingsley
(1973).
Data Extraction Technique
LANDSAT-1 MSS Frame 1017-16093 and one of its lakes, Lake Koshkonong
(56), will be used to illustrate the methodology used to extract the
digital information contained in a set of CCT's.
The extraction process commences with a change in the format of the
Goddard Space Flight Center (GSFC) CCT's and an expansion to 8-bit
mode giving a total of 256 digital number (DN) levels (0 to 255) of
optical intensity. The system is capable of handling the data
contained on two CCT's at a time, resulting in an initial output
consisting of one half of an LANDSAT-1 frame. An example of the left
half of Frame 1017-16903 is shown in Figure 18. The image has been
reconstructed using the IR2 DN values and a fiducial system has been
applied to the edges.
The section of Frame 1017-16093 containing Lake Koshkonong and a
portion of the surrounding terrain is extracted from the tapes in each
of the four MSS bands by supplying the computer with the appropriate
coordinates. Figure 19 shows the extracted section for Band 7 (IR2).
The histogram below the picture represented the full range of DN
values present within the scene on a scale of 0 (black) to 255 (white).
The prominent DN values, centered at approximately DN 128 and spreading
symmetrically, are related to the land features within the section.
The small cluster of DN values at the lower end of the DN scale is
associated with water features including Lake Koshkonong and the Rock
River.
In this study, the land DN values are of little interest and must be
eliminated, leaving just the water body values. Previous testing has
verified that the IR2 band is a good indicator of the areal extent of
surface water. As was noted in Figure 19, IR2 water-related intensity
values fall within the lower end of the DN scale and are essentially
isolated from the land IR2 DN values. This characteristic permits the
development of a binary mask which is used to eliminate the land image.
The mask is created by setting all intensity values at and below a
specific numeric value equal to a value of one and setting all of the
remaining DN values equal to zero. Each spatially equivalent Lake
Koshkonong pixel within each band (green, red, IR1, and IR2) is
multiplied by the IR2 binary mask. The net effect is to "zero out"
60
-------
Figure 18. Left half of MSS Frame 1017-16093 reproduced using IR2
DN values. A fiducial system has been imposed along the
edge of the image to aid in the determination of feature
locations on the basis of line and picture element counts,
Lake Koshkonong is the large dark object in the upper
center of the photograph.
61
-------
mnmmmmmmmummmmmmmmmmmmmmmmmmmmmmmm
Figure 19. An extracted section of LANDSAT-1 MSS Frame 1017-16093. Lake
Koshkonong, the Rock River, several small lakes and land
features are evident. A histogram of the section, seen below
the image, displays the range of DN values contained within
the section. Water DN values cluster toward the lower end
of the scale.
62
-------
the unwanted background, leaving just the water features. Figures 20
and 21 were created by multiplication with a binary mask produced by
setting all IR2 DN values of 28 or lower to a value of one. Images
constructed from the IR1 and IR2 would be similar, showing only
variations in the spectral signature for the water, and therefore in
their pixel DN histograms.
A final cleanup is done by upgrading the binary mask prior to multiply-
ing all of the bands. This is done by eliminating all of the smaller
water bodies, streams, and swamp features which may be present. After
the final cleanup and examination of a test multiplication, the mask
is used on all of the bands for Lake Koshkonong. The pictorial
expression of the final cleanup for the four spectral bands is depicted
in Figure 22.
The concatenation technique is a very convenient method for summarizing
the area! aspects of data extracted from the CCT's. Unlike line-printer
copy, areal coverage afforded for several different lakes can be
depicted in one simple photograph (e_._g_., Figure 23). Differences among
the lakes in Figure 23 are not apparent because the range of IRs DN
values is very small, and no effort has been made to enhance the images.
The VICAR photographic products displayed up to this point serve mainly
to aid the investigator in the extraction of digital data from the
CCT's. Typical numeric output from the system includes pixel counts,
DN means and standard deviations for each of the four MSS bands, and
histograms of the DN distributions. Pertinent data extracted from the
example, Lake Koshkonong, are listed in Table 11.
TABLE 11. DESCRIPTIVE STATISTICS OF LAKE KOSHKONONG MSS DATA
EXTRACTED FROM FRAME 1017-16093 CCT'S
LANDSAT-1 MSS BANDS
Green Red IR1 IR2
DN Mean
DN Standard
Deviation
Pixel Count
46.85
1.814
9,247
31.15
2.446
9,247
27.95
2.933
9,247
10.08
2.890
9,247
The DN values are used in determining correlations with ground (i_.;e.,
water) truth, developing regression models for selected lake trophic
indicators and trophic state, and for image enhancement.
63
-------
Figure 20. First-stage MSS Band 4 (6RN) cleanup picture of Lake Kosh-
konong. This picture was created by multiplying the Band
Four scene equivalent of Figure 19 by the binary mask. All
pixels corresponding spatially to IR2 pixels with DN values
greater than 28 have been multiplied by zero, and thereby
removed from the scene. The Rock River and miscellaneous
water bodies still have to be eliminated.
64
-------
Figure 21. First-stage MSS Band 5 (RED) cleanup picture of Lake Kosh-
konong. This picture was created by multiplying the Band
Five scene equivalent of Figure 19 by the binary mask. All
pixels corresponding spatially to IR2 pixels with DN values
greater than 28 have been multiplied by zero, and thereby
removed from the scene.
65
-------
Figure 22. Four band (GRN, RED, IR1, IR2) concatenation of Lake
Koshkonong after final cleanup. All of the unwanted land
and water body pixels have been eliminated.
66
-------
..•
SERIAL
NUMBER
Kegonsa
Rock
Koshkonong
Lac La Belle
Qconomowoc
Okauchee
Pine
Naqawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Figure 23. IR2 concatenation of 15 lakes extracted from LANDSAT-1 MSS
Frame 1017-16093. The lake images are all reproduced at
the same scale. However, they have not been skewed to
correct the geometric distortion induced by the earth's
rotation.
67
-------
DIGITAL IMAGE ENHANCEMENT TECHNIQUES
Up to this point, the VICAR photographic output has served as an aid in
the extraction of lake MSS DN values. However, the system can also
enhance the digital images to bring out latent features. Two methods
commonly employed are linear contrast stretching and a technique which
utilizes color ratios. The techniques are briefly explained in this
section along with several examples of the photographic output.
Color Ratio Technique
The color ratio technique consists of dividing the MSS DN values of one
LANDSAT-1 band by another LANDSAT-1 band, pixel by pixel (i.e.., data
point by data point). Water colors which are difficult to discriminate
generally have spectral reflectivity curves with similar but not identi-
cal slopes (Blackwell and Boland, 1974). The difference in the curve
slopes will form the basis of the discrimination rather than absolute
reflectivities. Application of the digital color ratio technique
between selected bands will normalize the data by removing the common
brightness components and will tend to emphasize differences due to
slope.
The digital image of Lake Koshkonong (Frame 1017-16093) was enhanced
using the ratio technique and subsequently reproduced in the form of a
concatentation (Figure 24). Compare the enhanced images with Figure 22.
It is readily apparent that considerable differences in lake color
exist, particularly in the GRNRED, GRNIR1, and REDIR1 images. These
ratios appear to merit further consideration. Additional information
concerning ratio techniques and their applications is found in
Billingsley and Goetz (1973).
Linear Contrast Stretching Technique
Another technique for detecting MSS color differences involves the
enhancement of digital contrast of selected MSS bands through a
"stretching" process. As is evident in Figure 18-23, the images of
Lake Koshkonong are of very low contrast and essentially featureless.
ihis is a consequence of the restricted DN value-range present for the
lake in all MSS bands. An expansion of the lake's DN ranges to fill
the entire black to white range (0 to 255) by linear expansion will
enhance the contrast and bring out latent features. Stretching in the
digital domain has the advantage that it does not suffer from the toe
and shoulder saturations encountered with photographic stretching
(Billingsley and Goetz, 1973).
The contrast stretch was performed on Lake Koshkonong data extracted
from two LANDSAT-1 MSS frames (1017-16093, 9 August 1972; 1036-16152,
28 August 1972). The results are presented as Figures 25 and 26. The
very strong banding effect is an artifact created by an inherent defect
68
-------
fn the MSS system. A comparison of the lake images suggests that the
MSS is capable of monitoring spatial-temporal changes in lakes. A
scarcity of ground truth precludes a full interpretation of the physical
significance of the patterns. This is one area in need of additional
investigation.
THU HDV 155
004600 JPL/IPL
Figure 24. Lake Koshkonong image enhancement using six LANDSAT-1 MSS
ratios. The ratios are, scanning from the top row left,
GRNRED, GRNIR1, 6RNIR2, REDIR1, REDIR2, IR1IR2. Differences
in lake color are particularly evident among the GRNRED,
GRNIR1, and REDIR1 ratios. The MSS data were extracted
from Frames 1017-16093 (9 August 1972). Compare with Figure
22.
69
-------
Figure 25. Contrast stretched images of Lake Koshkonong. Frame 1017-
16093 recorded by the LANDSAT-1 MSS on 9 August 1972.
Upper row, left to right: GRN, RED; lower row, left to
right: IR1, IR2. The banding or striping is an artifact.
70
-------
Figure 26. Contrast stretched images of Lake Koshkonong. Frame 1036-
16152 recorded by the LANDSAT-1 MSS on 28 August 1972.
Upper row, left to right: GRN, RED; lower row, left to
right: IR1, IR2. Compare with Figure 25.
71
-------
SECTION VI
LANDSAT-1 MSS-TROPHIC INDICATOR RELATIONSHIPS
This section is devoted to the evaluation of MSS data as a means of
estimating the magnitude of trophic state indicators including Secchi
disc transparency and chlorophyll a_. In addition, lake surface area
estimation is also examined,,
OPTICAL PROPERTIES OF PURE WATER AND NATURAL WATERS
It is readily apparent, even to the casual observer, particularly if
he or she is looking downward from an aircraft, that lakes differ in
color and brightness. Many investigations have been undertaken to
develop a comprehension of the processes which result in the observed
phenomena,, Although a detailed discussion of the interaction of electro^
magnetic energy with the components of the hydrosphere and atmosphere
is outside the scope of this report, a brief survey is essential to
gain some understanding of the principles which permit remote sensing
and yet constrain its use in the assessment of trophic indicator
magnitudes.
The interaction between electromagnetic energy and chemically pure
water has been studied by numerous investigators (e_.g_., Ewan, 1894;
Sawyer, 1931; Collins, 1925; James and Birge, 1938; Hulburt, 1945;
Raman, 1922; Dawson and Hulburt, 1937). The transmission of electro-
magnetic energy through a material medium is always accompanied
by the loss of some radiant energy by absorption. Some of the energy is
transformed into other forms (e_.c[., heat, chemical) or to some longer
wavelength of radiation (James and Birge, 1938). Pure water is very
transparent to violet, blue and green light. In the infrared region,
the extinction coefficient is high with a complementary low degree of
transmission (Table 12). The absorption spectral characteristics of
pure water can be modified greatly through the addition of dissolved
and particulate materials.
The absorption spectra of natural waters (e_.£., lake and ocean) have
been studied in detail by Jerlov (1968), Duntley (1963), Atkins and
Poole (1952), Birge and Juday (1929, 1930, 1931, 1932), and Juday and
Birge (1933), to mention a few. Hutchinson (1957) has summarized the
more important attempts to elucidate the interactions of light with
natural waters, particularly in regards to lakes.
The attenuation of electromagnetic radiation in lake waters is a conse-
quence of the relatively unselective effect of suspended particulate
materials and the highly selective effect of dissolved coloring matter,
72
-------
TABLE 12. OPTICAL PROPERTIES OF PURE WATER (ROOM TEMPERATURE)'
Wavelength Extinction Percentile Refractive
(nanometers) Coefficient Absorption Index
820 (infrared)
800
780
760
740
720
700
680 (red)
660
640
620 (orange)
600
580 (yellow)
560
540
520 (green)
500
480
460 (blue)
440
420
400 (violet)
380 (ultraviolet)
2.42
2.24
2.31
2.45
2.16
1.04
0.598
0.455
0.370
0.310
0.273
0.210
0.078
0.040
0.030
0.016
0.0075
0.0050
0.0054
0.0078
0.0088
0.0134
0.0255
91.1
89.4
90.1
91.4
88.5
64.5
45.0
36.6
31.0
26.6
23.5
19.0
7.0
3.9
3.0
1.6
0.77
0.52
0.52
0.70
0.92
1.63
2.10
1.329
1.331
1 .333
1.338
1.343
aAdapted from Hutchinson (1957).
usually of organic origin, on the electromagnetic spectrum. The dis-
solved matter absorbs strongly in the violet and blue wavelengths,
moderately in the middle wavelengths (e_.g_., green), and much less
strongly at longer wavelengths (Hutchinson, 1957). When the dissolved
materials are present in small quantities, the water will be most trans-
missive in the green wavelengths. Lake waters with large amounts of
dissolved substances will be more transmissive in the orange and red
wavelengths than in the shorter wavelengths. However, the transmission
of the red and orange light is still greater in pure water than in water
containing particulate and/or dissolved materials. As water trans-
parency diminishes, the detectable electromagnetic energy will be of
progressively longer wavelength, at increasingly shallower depths
(Hutchinson,.1957).
73
-------
Scherz, et al_. (1969) have investigated the total reflectance (surface
reflectance plus volume reflectance) curves of pure water and natural
waters under laboratory conditions using a spectrophotometer. The
resulting curve for distilled water is shown in Figure 27. They
reported that the addition of dissolved oxygen, nitrogen gases, and
salts (e.g.., NaCl, Na^CK, Na3P04-H20) had no apparent effect on the
reflection curve. However, water from lakes in the Madison (Wisconsin)
area had reflectance curves that both differ from the distilled water
curve and from each other. Scherz, et_ al_. suggest that the differences
can be attributed to the presence of different algal organisms. Filtra-
tion of the lake waters produced similar reflectance curves, though
different from that of pure water (Figure 28).
10
Ul
o
z
<
I-
o
UJ
a:
O.I
I
200
400 600 800 1,000 1,200
WAVELENGTH (NANOMETERS)
Figure 27. Reflectance curve for distilled water. Adapted from
Scherz, ejt a]_. (1969).
74
-------
10
o
LU
_l
u_
Ill
tt: I
iu 0.4
o
tr
UJ
o.
0.0
FILTERED
UNFILTERED
200 400 600 800 1,000 1,200
WAVELENGTH (NANOMETERS)
H 4
o
UJ
u_
UJ
tc.
UJ
o
a:
LU
O.
0.4
0.0
— UNFILTERED
\ I LAKE \
^ KEGONSA
\
\
200 400 600 800 1,000 1,200
WAVELENGTH (NANOMETERS)
Figure 28. Reflection characteristics of filtered and unfiltered water
samples from two Wisconsin lakes in the area of Madison.
Adapted from Scherz, et^ aj_. (1969).
75
-------
The color of a lake is the color of the electromagnetic energy back-
scattered from the lake body to the sensor. Lake color ranges from
the blue of pure water through greenish blue, bluish green, pure green,
yellowish green, greenish yellow, yellow, yellow brown, and clear
brown (Hutchinson, 1957). It is a common practice to record lake color
on an empirical scale such as the so-called Forel-Ule color scale,
originally devised by Forel (1889) and subsequently modified and
extended by Ule (1892). Lake color need not be, and is usually not the
same as, the color of the lake water*.
Lakes which are blue in color lack appreciable quantities of humic
materials and colored materials in suspension (e_.g_., phytoplankton).
The bluer the lake color, the smaller the amount for free-floating
organisms contained in the water (Ruttner, 1963). Water with high
plankton content possess a characteristic yellow-green to yellow color.
The characteristic color may not be apparent due to masking by other
materials (e_.£., suspended sediments). Ruttner (1963) suggested that
A lake with very transparent and dark blue, blue-green, or
green water is always oligotrophic. On the other hand,
eutrophic lakes always have a relatively low transparency
and are yellow-green to yellow brown in color; but the
determination of these optical properties alone will not
establish the productivity type, for the turbidity can be
of inorganic origin, and the color can come from humic
substances.
Seston color, attributable to the reflection spectra of suspenoids of
microscopic or submicroscopic size, is often observed in highly
productive lakes. Lakes containing large quantities of suspended in-
organic matter (e_.cj_., silt) may acquire a characteristic seston color,
but in most cases the color is related to large concentrations of
phytoplanktonic organisms (Hutchinson, 1957).
Peripheral Effects
The character of the electromagnetic energy impinging on the remote
sensor, the LANDSAT-1 MSS in this case, has been shaped through inter-
action with numerous environmental phenomena.
The earth's atmosphere has a pronounced effect on the solar spectrum.
The spectral distribution of the sun's radiation at the outer edge of
the atmosphere and the normal energy distribution at the earth's surface
* Welch (1952) defines water color as "...those hues which are inherent
within the water itself, resulting from colloidal substances or sub-
stances in solution" (i_.e_., true color). The platinum-cobalt scale
(Hazen, 1892) has found favor in the United States for the determina-
tion of water color.
76
-------
are illustrated in Figure 29. Atmospheric conditions (e_.£., degree of
cloudiness; presence of fog, smoke, and dust; amount of water vapor)
affect the degree of insolation attenuation. Weather conditions
strongly affect the distribution of energy between sunlight and skylight
(Piech and Walker, 1971), contributing a degree of uncertainty in water
quality assessment through remotely sensed color measurements. Hulstrom
(1973) has pointed out the adverse impact that cloud bright spots can
have on remote sensing techniques which utilize reflected energy.
The degree of scattering and absorption imposed on the return signal from
water bodies is related to the atmospheric transmittance and can result
in changes in lake color when sensed at aircraft high flight and satel-
lite altitudes. The attenuated return signal is also contaminated by
electromagnetic radiation from the air column (path radiance). Rogers
and Peacock (1973) have reported that solar and atmospheric parameters
have a serious adverse impact on the radiometric fidelity of LANDSAT-1
data. Path radiance was found to account for 50 percent or more of the
signal received by the MSS when viewing water and some land masses.
The magnitude of the adverse atmospheric effects can be reduced, though
not completely eliminated, by using imagery or digital data collected
on clear, cloudless days. This is the approach used in this investiga-
tion.
The LANDSAT-1 spacecraft passes over the same point on the earth at
essentially the same local time every 18 days. However, even though
the time of passover will remain essentially the same throughout the
year, solar elevation angle changes (Figure 30) cause variations in the
lighting conditions under which the MSS data are obtained,, The changes
are due primarily to the north or south seasonal motion of the sun
(NASA, 1972). Changes in solar elevation angle produce changes in the
average scene irradiance as seen by the sensor from space. The change
in irradiance is influenced both by the change in the intrinsic reflec-
tance of the ground scene and by the change in atmospheric backscatter
(path radiance). The actual effect of changing solar elevation angle
on a given scene is very dependent on the scene itself (NASA, 1972). For
example, the intrinsic reflectance of sand is significantly more sensi-
tive to changing solar elevation angle than are most types of vegetation
(NASA, 1972). The effects of changing solar elevation angle are of
particular importance when comparing scenes taken under significantly
different angles. The use of color ratios in lieu of raw data values
may be of value in reducing the magnitude of the solar angle induced
effects by normalizing the brightness components. This is the approach
used in this investigation.
A portion of the radiation impinging on the lake surface will be
reflected. The percentage of surface-reflected energy is a strong
function of the angle of incidence (Figure 31). Surface roughness is
77
-------
00
SOLAR ENERGY DISTRIBUTION
OUTSIDE ATMOSPHERE
APPROXIMATE DISTRIBUTION IN
DIRECT SOLAR RADIATION
AT GROUND
(WAVELENGTH IN NANOMETERS)
0
200 400
ULTRAVIOLET
600 800 1000 1200 1400
VISIBLE NEAR INFRARED
1600 1800
Figure 29. Spectral distribution of solar energy. Adapted from
Hutchinson (1957).
-------
SOLAR
ZENITH
ANGLE
SOLAR
ELEVATION
ANGLE
TANGENT
PLANE
SUBSATELLITE
POINT
Figure 30. Solar zenith and solar elevation relationship,.
Source NASA (1972).
known to have an effect on the percentages of light
refracted at the interface (Jerlov, 1968). However,
surface is negligible in estimating total radiation
body when the solar elevation angle is greater than
(Hutchinson, 1957).
reflected and
the effect of
entering a water
15 degrees
79
-------
100
10 20 30 40 50 60 70 80
ANGLE OF INCIDENCE MEASURED FROM
NORMAL
Figure 31. Percentage reflectance of the air-water interface as a
function of the angle of incidence measured from normal
direction. Values are for unpolarized light only. Source,
Piech and Walker (1971).
80
-------
The light reflected from the interface is composed of diffuse light
from the sky (skylight) and specularly reflected sunlight. Specular
reflection areas contained in a scene are of little value in most water
studies, the possible exception being the determination of surface
roughness. The specularly reflected radiation exceeds, by several
orders of magnitude, the reflected energy emanating from beneath the
water surface (Curran, 1972). Surface reflected skylight, containing
no water quality color information, can compose from 10 percent of the
return signal on a clear day to 50 percent on a cloudy day (Piech and
Walker, 1971). The surface reflected skylight not only increases the
apparent reflectance from the water body (volume reflectance), but also
affects the shape of the reflectance curve.
An electromagnetic wave impinging on the surface of a lake decomposes
into two waves, one of which is refracted, proceeding into the aquatic
medium and a second wave which is reflected back to the atmosphere
(Jerlov, 1968). The wave entering the water is refracted as it passes
through the air-water interface according to Snell's Law which may be
expressed as:
n = sin i/si'n r
where (i) is the angle of incidence, (r) is the angle of refraction, and
(n) is the refractive index, which for water is approximately 1.33 (see
Table 12).
Most of the electromagnetic energy entering a lake is attenuated through
the process of absorption. Although only a small percentage (less than
3 percent; Davis, 1941) of the incident energy is backscattered from the
lake water volume, this light (volume reflectance) is the focus of
interest in remote sensing of water quality investigations. Its spectral
characteristics have been shaped by the materials found in the lake
waters (dissolved and suspended materials, plankton, aquatic macrophytes,
and air bubbles).
The lake bottom characteristics (color and composition) will also affect
the intensity and/or the spectrum of the volume reflectance in settings
where water transparency permits the reflection of a significant amount
of radiation from the bottom materials. In studies involving the esti-
mation of water depth or the mapping of bottom features, it is essential
that the lake bottom be "seen" by the sensor. Bottom effects are
capitalized upon and put to a beneficial use.
However, in this investigation, bottom effects are considered to be an
undesirable peripheral effect. A sensor with the capabilities of the
LANDSAT-1 MSS is not able to "see" much deeper into a lake than Secchi
disc depth. The Secchi disc transparency of the selected NES lakes is,
in most cases, relatively small when compared to the mean depth of each
lake The assumption is made, as a first approximation, that bottom
81
-------
effect is relatively insignificant when considering each of the selected
lakes as an entity.
It is evident that many factors influence the intensity and spectral
characteristics of the electromagnetic radiation which is collected by
the sensor. A peripheral effect, desirable in one study, may have a
deleterious effect in another study. Absolute quantification of re-
motely sensed phenomena requires that all of the effects be accounted
for in the return signal. Although the approach used here to reduce the
magnitude of the undesirable peripheral effects might be criticized as
simplistic or naive, it does serve as a starting point in the investiga-
tion of lake color-trophic indicator relationships using satellite-borne
sensors. Although the results may be of a semi-quantitative nature,
general trends in the data may be of value in lake classification
studies.
RELEVANT REMOTE SENSING LITERATURE
A wealth of literature exists relating to the theoretical, applications,
and instrumentation aspects of remotely sensed water quality. The
advantages and limitations of remotely assessing water quality have
been discussed by many investigators (Robinove, 1965; Horn, 1968;
Kolipinski and Niger, 1968; Clarke, 1969; Fortunatov, 1957; Kiefer and
Scherz, 1970, 1971; Conrod and Rottweiler, 1971; Gramms and Boyle, 1971;
Scherz, 1971; Clapp, 1972; Wezernak and Polcyn, 1972; Uolwell, 1973).
Wezernak and Polcyn (1972) examine the question of making eutrophication
assessments from the standpoint of current remote sensing technology.
They suggest, that of some 16 factors which often serve as measures of
eutrophication (Table 13), several can be remotely sensed using opera-
tional and near-operational systems. The parameters include chloro-
phyll, colored water masses, suspended solids, transparency, aquatic
macrophytes, and algal blooms. In addition, remote sensing technology
can also provide an economic method of obtaining morphometric and land
use information (e_.^_-> shoreline development, lake surface area, and
cultural impact) which has a direct bearing on the trophic state of a
water body.
Most of the water-oriented investigations have focused on the oceanic
environment and point source pollution of fresh waters using aerial
photography (black and white, color, black and white infrared, color
infrared, multispectral). Investigations utilizing airborne multi-
spectral scanners and lasers are becoming increasingly common. Many
aspects of water quality have been examined using remote sensors
mounted on non-satellite platforms; pertinent aspects are discussed
below.
Remotely determined estimates of chlorophyll levels in natural waters
(in situ) have been made by several investigators (Clarke, et al_., 1969,
82
-------
TABLE 13. INDICES COMMONLY USED TO ASSESS EUTROPHICATION3
Standing crop of algae and aquatic plants'3
Amount of suspended solids^
Volume of algae
Chlorophyll levels'-1
Number of algal blooms^
Transparency^
Plant regression'3
Photosynthesis
Primary production
Aquatic plant nutrient content
Hypolimnetic oxygen concentrations
Sediment composition
Dissolved solids
Conductivity
Nutrient concentrations
Cation ratio (Na + K) / (Mg + Ca)
? Adapted from Wezernak and Polcyn (1972).
Index can be remotely sensed using operational
or near-operational sensors.
1970; Mueller, 1972; Curran, 1972; White, 1969; Atwell and Thomann,
1972; Arvensen, et_ aJL , 1971; Bressette and Lear, 1973). Curran
(1972) reported a strong correlation between wavelength-dependent
albedo ratios, made from measurements collected by high-flight aircraft,
and phytoplankton chlorophyll concentrations, from satellite altitudes,
to an uncertainty of 0.1 mg/m3. Blackwell and Billingsley (1973) have
utilized multispectral photography in conjunction with digital computer
enhancement techniques to detect algae. Crew (1973) has mapped and
identified an algal species in Clear Lake (California) employing an
airborne multispectral scanner.
Aerial photography, utilizing color and color-infrared films, is a
relatively simple technique that has demonstrated utility in mapping
aquatic macrophytes (Kolipinski and Higer, 1968; Kiefer and Scherz,
1971) and shallow water benthos (Kelly and Conrod, 1969). Photographs
are used routinely to map wetlands (Anderson and Wobber, 1973). It is
an accepted practice to use aerial photographs in the delineation and
enumeration of lakes (Minnesota Department of Conservation, 1968).
Remote sensors are employed to estimate water turbidity (Crew, 1973;
Schmer, et al_., 1972, Williams and Samol, 1968). Some investigators
(Brown, et al_., 1972; Specht, et al_., 1973) have used remote sensors to
83
-------
estimate water depth and bottom topography. Dybdahl (1973) has demon-
strated an airborne remote sensing technique for the determination of
dissolved oxygen levels in fresh water.
Relevant LANDSAT-1 Investigations
Prior to the insertion of LANDSAT-1 into its near-polar orbit, investi-
gators interested in assessing and monitoring the earth's natural
resources from satellite altitudes were restricted to using poorly
suited satellite-borne sensors. The spacecraft were largely equipped
with sensors designed for studies oriented toward the atmosphere or,
in some cases, the ocean.
The length of the list of NASA-designated LANDSAT-1 principal investi-
gators -- some 300 are found on it (Data Users Handbook, Appendix M,
February, 1972) -- suggests that there is a great interest in assessing
the advantages and limitations of remotely sensing the earth's resources
from satellite altitudes. The major focus, as evidenced by project
titles, is on the terrestrial environment; less than 10 percent of the
investigations relate to water quality and the majority of these
concentrate on the oceanic environment and the Laurentian Great Lakes.
Some of the water-related investigations bear mentioning.
Wezernak and Polcyn (1972) have detected a municipal-industrial waste
disposal in the New York Bight. Lind and Henson (1973) and Lind (1973)
have discovered and documented a pollution plume emanating from a
mill located on the shore of Lake Champlain. It is apparent from a
literature review that the detection of large turbidity plumes and
turbidity related patterns is easily accomplished using the LANDSAT-1
multispectral scanner-generated imagery (e_.c[., Watanabe, 1973; Klemas,
1973; Carlson, 1973; Wright, Sharma, and Burbank, 1973; Pluhowski,
1973; Coker, Niger, and Goodwin, 1973; Kritikos, Yorinks, and Smith,
1973).
Enhanced and density-contoured imagery appears to permit water depth
estimation to a depth of at least five meters in the clear waters around
the Bahama Islands (Ross, 1973). Polcyn and Lyzenga (1973), using
digital processing techniques on an LANDSAT-1 MSS frame taken over the
Bahama Islands, have mapped shallow water features and calculated water
depths to five meters. The waters in the area are known for their
clarity and are conducive to remote sensor mapping activities.
Hidalgo, et al_. (1973) have identified large mats of duckweeds
(LemnaceaeT on Lake Pontchartrain and surrounding bayous and swamps
in southeastern Louisiana. Seasonal changes have been detected. Strong
(1973) has noted algal blooms in Utah Lake (Utah) and in Lake Erie.
Barr (1973) has mapped a total of 2,272 water bodies in Saline County
(Kansas) where a topographic map (1955) indicates 1,056. A preliminary
comparison of imagery and maps indicates that bodies larger than four
84
-------
hectares (10 acres) are consistently detectable. Most water bodies
between about two to four hectares are usually detectable; water
areas less than two hectares are occasionally resolved. Erb (1973),
using nine-inch-square transparencies, has determined that ponds as
small as one hectare are detectable within forested areas. The
initial findings of Chase, Reed, and Smith (1973) indicate that water
bodies of about 0.5 hectares are detectable under fair conditions
(haze and 70 percent cloud cover); distortion of lake size, shape, and
orientation is minimal. The benefits of using LANDSAT-1 imagery to
enumerate lakes have been demonstrated by Reeves (1973) and Work, et
al_. (1973). ~~
Yarger, e_t al_. (1973), using electronically sliced imagery, have found
a good correlation between suspended load (predominantly composed of
inorganic materials) and film densities of two federal reservoirs in
Kansas. Bowker, ert a]_. (1973) reported that there appears to be a
positive correlation between particulate count and chlorophyll level in
Lower Chesapeake Bay. A rough correlation exists between suspended
sediments and MSS imagery, but no correlation is apparent for the
chlorophyll bearing portion of the load. Szekielda and Curran (1972)
indicate that a correlation exists between chlorophyll and LANDSAT-1
MSS imagery (green band) in the general area of St. John's River
estuary in Florida.
Rogers and Smith (1973), reporting on their investigation of six lakes
in Michigan, suggested that deep water and shallow water can be separated
by a trained photo-interpreter using reflectance printout gray-scales
or a cathode tube monitor. Lakes were best discriminated in band seven
(IR2) due to the strong contrast between land and water (Rogers and
Smith, 1973). Classifying lake eutrophication on the basis of algal
scum or macrophytes in shallow water appears to be a "straight-forward"
matter with LANDSAT-1 MSS data. Estimations of water depth may be
possible in some lake areas where there are extensive shallows and the
water is clear and unencumbered with vegetation (early spring).
TROPHIC INDICATOR ESTIMATION
Although the LANDSAT-1 provides 18-day cyclic coverage, a point stressed
in many of the reports and articles written on the spacecraft, obtaining
good coverage in the study area was (and still is) a major problem.
This is a problem shared with many other investigators. On numerous
dates of coverage, cloud cover was excessive (greater than 10 percent)
and on several occasions, when the weather conditions were conducive to
monitoring the lakes, one or more of the MSS bands were either missing
or rated as poor. The formation of ice cover on the study lakes,
lasting for several months, further reduced the opportunity to obtain
repetitive coverage.
85
-------
A search of the LANDSAT-1 MSS imagery and CCT's for the study area
(Minnesota, Wisconsin, Michigan, New York) for the period from August,
1972 through July, 1973, resulted in the selection of the frames found
in Table 14. The lakes which were examined in this investigation are
listed in Table 15. Several additional frames were not included
because they contain less than three lakes sampled under the National
Eutrophication Survey Program. Frames from the 1973 calendar year were
not selected if LANDSAT-1 MSS coverage was not available for the 1972
sampling year.
TABLE 14. LANDSAT-1 MSS FRAMES
Frame
Number
1017-16091
1017-16093
1036-16152
1323-16094
1323-16100
1359-16091
1359-16094
1022-16373
1075-16321
1077-16431
1309-16325
1345-16322
1346-16381
1027-15233
1080-15180
Date
9 August 1972
9 August 1972
28 August 1972
11 June 1973
11 June 1973
17 July 1973
17 July 1973
14 August 1972
6 October 1972
8 October 1972
28 May 1973
3 July 1973
4 July 1973
19 August 1972
11 October 1972
Area
Eastern Wisconsin
Southeastern Wisconsin
South central Wisconsin
Eastern Wisconsin
Southeastern Wisconsin
Eastern Wisconsin
Southeastern Wisconsin
Central Minnesota
Central Minnesota
West central Minnesota
Central Minnesota
East central Minnesota
Central Minnesota
Northwestern New York
Northwestern New York
Number of
Lakes
5b
15b
6
4b
23b
4b
18b
12
15
10
13
14
8
7
5
Lakes extracted from frames recorded on the same flightline and data
^are pooled to increase sample size.
Some of the lakes appear in both frames; e_.£., Lake Winnebago is
often split between two frames.
86
-------
TABLE 15. DATES OF LANDSAT-1 COVERAGE
LANDSAT-1 Coverage Data
Lake Name and
1972
1973
Serial Number
Shawano
Butte des Morts
Poygan
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac la Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Mendota
Mononaa
Waubesaa
Darl ing
Carlos
Le Homme Dieu
Minnewaska
Nest
Green
Wagonga
Clearwater
Maple
Cokato
Buffalo
Carrigan
0)
cr> en 01 cr> -i-> 4-> +J >>c
3333OOO(O3
eteteteCOOOIEI'-a
cn ^- cr> oo vo oo i— co r—
i— i— CM i — CM i—
46
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
108
109
110
16
17
18
19
20
21
22
23
24
25
26
27
X
X
X
X
X
X X
X X
X X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X
x x
x x
x x
x x
x x
X X
x x
x x
x x
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
333
•~D I~D rD
co •* r^.
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X X
X X
87
-------
TABLE 15. DATES OF LANDSAT-1 COVERAGE (continued)
LANDSAT-1 Coverage Data
1972 1973
Lake Name and
Serial Number o> o> o> o» -u +j ••-> >,E.^^
3333OOO(O33:
c£ «C eCeCOOOSCrSO1":
OH ^- Ot CO IO 00 i— OOi— CO «3
1—1— CM i— CM i—
Silver
Minnetonka
Forest
White Bear
St. Croix
Spring
Pep in
Madison
Sakatah
Winona
Trace
Calhoun
Big Stone
Zumbro
Cottonwood
Conesus
Canandaigua
Keuka
Seneca
Cayuga
Owasco
Cross
One i da
Canadarago
28
29
30
31
32
33
34
35
36
101
102
103
104
105
111
91
92
93
94
95
96
97
106
107
X X
X X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X X
X X
X
X
X
X
X
X
X
X
X
X
X X
X
X
X
X
X
X
X
X
X
This lake fell outside the scope of the National Eutrophication Survey
Program, but was included here because it is a well-known lake on
which considerable eutrophication research has been conducted.
88
-------
Every effort was made to get MSS coverage which was concurrent with the
collection of the ground truth by the helicopter-borne field teams. This
goal was achieved in the case of only one frame (1080-151800). Unfortu-
nately the number of NES sampled lakes contained in the frame is too small
to incorporate into a regression model. The dates on which ground truth
was collected from the study lakes are found in Appendix B.
Ideally, the estimation of the magnitude of trophic state indicators should
be done using concurrent data to derive the maximum benefit. However, in
this investigation it was necessary to use what may be called "near concur-
rent" ground truth which was collected several days before or after satel-
lite overflight. Nevertheless, models developed from such a temporal
arrangement are of some value in illustrating general relationships exist-
ing between the MSS data and ground truth.
Estimations of the magnitude of two trophic state indicators (chlorophyll
a. and Secchi disc transparency), using LANDSAT-1 MSS data and NES collected
ground truth, are demonstrated in the remainder of this section using two
of the frames in Table 14. The frames, 1017-16091 and 1017-16093, are
treated as one frame to increase the lake sample size. The frames are in
juxtaposition on the same flightline, were recorded on the same day, and
prior to Goddard Space Flight Center Processing, were elements in a
continuous strip. The frames were selected on the basis of temporal
proximity to NES ground truth dates, the quality of the MSS data (good
MSS band ratings, little cloud cover and haze), and the presence of a re-
latively large number of NES lakes (N=20). An examination of Table 15
indicates that coverage is available for the same 20 lakes on two other
occasions.
MSS frames from 12 different dates and three states (Minnesota, Wisconsin,
and New York) are examined in this report. The fragmentary LANDSAT-1
coverage, very evident in Table 15, makes it difficult to give a coherent
demonstration of the MSS's capabilities and limitations. In an attempt
to reduce the magnitude of the problem, the author has taken the liberty
of focusing on the Wisconsin frames and relegating the other frames to
Appendix D. The appendix is divided into a series of subappendices (one
for each LANDSAT-1 date) which contain descriptive statistics of the MSS
data, LANDSAT-1 MSS estimates of lake area, area ratios, lake concate-
nations, three-dimensional color ratio models, and regression models.
Data reduction was accomplished on the Oregon State University CDC 3300
computer using the Statistical Interactive Programming System (SIPS). The
regression models were developed using the backward selection procedure
(Guthrie, e_t a]_., 1973). The backward method was selected largely on the
basis of its intuitive appeal.
89
-------
LANDSAT-1 Lake Area Estimation
A straightforward method of estimating the area of an extracted lake
involves a summation of the pixel counts in one band (e_.£., IR2) and
subsequent multiplication by a conversion factor of 0.48 (1 pixel =
0.48 ha = 1.18 ac) to get the area in hectares. The conversion factor
used here was obtained from R. J. Blackwell (personal communication,
1973). A comparison of lake area estimations from other sources with
LANDSAT-1-derived area values acts as a check to verify that the pixels
and their respective DN's are those of the water body under scrutiny.
A topographic area/LANDSAT-1 area ratio greater than one indicates that
the extracted image is smaller than the area covered by the MSS lake
image. A ratio less than one suggests that the extracted image includes
wet lands and/or land features. Underestimation of lake size is pre-
ferable to overestimation (inclusion of wetland and/or land) in this
study. The surface areas of the 20 lakes extracted from Frames 1017-
16091 and 1017-16093 are generally within 10% of values derived from
U.S.G.S. topographic sheets and publications of the Wisconsin Department
of Natural Resources (see Table 16).
Discrepancies between LANDSAT-1 and topographic sheet-derived area
estimates may be a consequence of: changes in the lake area since the
topographic sheet was constructed; inaccurate delineation of the shore-
line on aerial photographs or during the preparation of the sheet;
inclusion of wetlands or exclusion of portions of the lake during the
preparation of the IR2 binary mask; or failure to use a correct
conversion factor.
The estimated lake areas using LANDSAT-1 MSS pixel counts are in good
agreement for most of the lakes in Table 16. The very large error for
Middle Lake is due to the contiguous nature of the Lauderdale Lakes.
The computer extraction process resulted in a sample taken from the
area mapped as Middle Lake. Some of the lakes, particularly those which
occupy shallow basins, are known to fluctuate greatly in surface area
and topographic sheets do not account for the variation.
Adding the 20 lake areas derived from the topographic sources and then
comparing the resultant value with the composite LANDSAT-1 MSS lake area
value, excluding Middle Lake, results in a ratio of 1.016:1.000. A
visual examination of the area ratios for the other frames (see the
subappendices of Appendix D) indicates that the MSS can be used to give
good estimates of lake surface area when a DN value of 28 is used as
the "cutoff" point in extracting the lakes from their terrestrial matrix.
The lake area estimation capabilities of the MSS are of value, not only
in the study of lakes with established areas, but also in geographic
regions for which there is no accurate topographic or aerial photo-
graphic coverage.
The photographic products (concatenations), fabricated by the VICAR
system, are of immense value in extracting the lake "images" stored on
90
-------
TABLE 16. AREAL ASPECTS OF 20 NES-SAMPLED LAKES EXTRACTED FROM
LANDSAT-1 MSS FRAMES 1017-16091 and 1017-16093.
Lake Name
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac la Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
Pixel
Count
9,177
7,395
114,186
6,613
5,162
2,675
1,009
9,247
944
652
871
568
751
1,996
689
309
117
1,541
703
4,543
LANDSAT-1 Lake
Area (ha)
4,382.5
3,531.5
54,529.1
3,158.0
2,465.1
1,277.4
481.8
4,415.9
450.8
311.4
415.9
271.3
358.6
953.2
329.0
147.6
55.9
735.9
335.7
2,169.5
Map Lake Map Area:
Area (ha) LANDSAT-1 Area Ratio
4,448.5
3,584.4
55,730.4
2,972.9
2,671.0
1,099.2
554.8
4,241.3
452.1
317.7
450.8
284.5
415.2
1,008.9
449.8
160.3
104.8
717.9
383.0
2,129.5
1.015
1.015
1.022
0.942
1.084
0.861
1.152
0.961
1.003
1.015
1.085
1.049
1.155
1.058
1.367
1.086.
1.876a
0.976
1.151
0.982
aMiddle Lake is contiguous with Mill Lake and Green Lake; the lakes are commonly
referred to as the Lauderdale Lakes. The extraction process resulted in a sample
of the area mapped as Middle Lake. The Green Lake of the Lauderdale Lakes is not
to be confused with Green Lake (51).
-------
the CCT's. They permit the investigator to examine the extracted lake
for geometric fidelity, truncation of the water body, and for the
inclusion of other water bodies. The photographic products are superior
to output in the form of line printer copy in studies relating to the
area! aspects of water bodies.
Secchi Disc Transparency Estimation
Although it is beyond the capabilities of the MSS to directly measure
chemical indicators, its area! and spectral resolution permit the
detection of phenomena related to eutrophication (such as Secchi disc
transparency, chlorophyll a_) together with some degree of quantification.
An R-mode Pearson product-moment correlation analysis was made using the
MSS data and ground truth collected from the 20 NES lakes in Frames
1017-16091 and 1017-16093. The correlations, based on data means and
transformed means, are found in Table 17.
TABLE 17. CORRELATIONS BETWEEN GROUND TRUTH AND LANDSAT-1
MSS DATA (COLORS AND COLOR RATIOS) FOR 20 LAKES
IN FRAMES 1017-16091 AND 16093.
PCI
CHLA
LNCHLA
SECCHI
LNSECCHI
GRN
RED
IR1
IR2
GRNRED
GRNIR1
GRNIR2
REDIR1
REDIR2
IR1IR2
0.518
0.722
0.807
0.589
-0.823
-0.806
-0.470
-0.544
-0.026
0.516
0.812
0.888
0.899
0.680
-0.821
-0.777
-0.442
-0.476
0.028
0.522
0.718
0.860
0.886
0.696
-0.886
-0.838
-0.521
-0.505
-0.017
0.474
-0.623
-0.788
-0.741
-0.492
0.865
0.685
0.357
0.274
-0.156
-0.527
-0.662
-0.857
-0.866
-0.642
0.919
0.803
0.476
0.430
-0.042
-0.507
Several multiple regression models were developed to predict Secchi disc
transparency using MSS colors and color ratios as the independent
variables. The use of color ratios as independent variables has appeal
because the use of ratios tends to normalize the data by removing the
brightness components. The best model, as measured by the magnitude
of its coefficient of multiple determination (R2) and standard error
of estimate, for estimating Secchi disc transparency is:
LNSECCHI = 0.784 + 2.638 GRNIR1 - 3.731 REDIR1 - 0.754 IR1IR2
92
-------
The model explains approximately 87 percent of the variance about the
mean (Table 18). The observed and predicted Secchi disc transparency
values for the 20 lakes are given in Table 19. Although the number of
observations is limited (N = 20) and the ground truth is sparse, it is
apparent that the LANDSAT-1 MSS can be used to estimate Secchi disc
transparency in freshwater lakes.
The most glaring disparity occurs with Middle Lake. Several factors
may individually or collectively account for the large residual, includ-
ing: bottom effects, presence of extensive beds of aquatic macrophytes,
the overlap of lake pixels onto land, or some still unsuspected factor.
Middle Lake has good water clarity, is relatively shallow, and does have
a problem with submerged aquatic macrophytes.
An effort was made to estimate the Secchi disc transparency of 11
Minnesota lakes found in Frame 1022-16373. The regression model, found
in subappendix D5, must be viewed with caution because the ground truth
was collected approximately two and one-half weeks after the date of
LANDSAT-1 coverage. Lakes are dynamic and significant changes may have
occurred during the intervening period.
Chlorophyll a_ Estimation
Bressette and Lear (1973), using an infrared photographic technique,
have detected algal blooms (primarily Anacystis) in the "salt wedge"
area of the Potomac River near Maryland Point. The inherent character-
istics of the spectral reflectance curve for chlorophyll-bearing plants
(Figure 32) were used to advantage. The reflectance of the chlorophyll-
bearing plants varies greatly as illustrated by the curve with an abrupt
increase at about 700 nanometers. Although the water tends to attenuate
the infrared energy in a relatively short distance, the magnitude of the
plant reflectance in the near-infrared allows the detection of
chlorophyll-bearing plants on or near the water surface. Bressett and
Lear (1973) proposed that "it is also probable that the magnitude of the
reflected solar energy in the NIR will also depend upon the concentration
of phytoplankton in the water..." and demonstrated their infrared tech-
nique for the mapping of chlorophyll ^concentrations.
An examination of the correlations between the MSS data (colors and color
ratios) and the chlorophyll a_ data from the 20'NES-sampled lakes [Table
17) suggests that it may be possible to estimate mean chlorophyll a*
levels using the MSS data.
A multiple regression analysis yielded the model:
LNCHLA = 6.905 - 3.662 GRNIR1 + 0.636 GRNIR2
* Mean value for each lake as determined from ground truth acquired on
a specific date.
93
-------
TABLE 18. ANALYSIS OF VARIANCE TABLE OF A REGRESSION MODEL
FOR THE PREDICTION OF SECCHI DISC TRANSPARENCY3
Source
Analysis of Variance
df Sum of Squares Mean Square Calculated F
Total (corrected) 19
Regression 3
Residual 16
16.340
14.152
2.188
0.860
4.717
0.137
34.431
The model was developed using MSS data from Frames 1017-16091 and
1017-16093.
TABLE 19. SECCHI DISC TRANSPARENCY RESIDUALS3
Lake Name
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac la Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
SECCHI
(m)
0.46
0.52
0.71
5.69
0.38
0.99
2.39
0.32
1.91
3.51
1.83
3.05
0.91
1.82
0.57
2.13
5.18
0.58
0.41
3.37
SECCHI
~ (m)
0.55
0.48
0.51
4.87
0.58
0.55
2.68
0.35
1.42
3.27
0.64
4.85
1.03
1.18
0.70
1.97
2.57
1.07
0.55
4.53
SECCHI-SECCHI
(m)
-0.09
0.04
0.20
0.82
-0.20
0.44
-0.29
-0.03
0.49
0.24
1.19
-1.80
-0.12
0.64
-0.13
0.26
2.61
-0.49
-0.14
-1.16
The model was developed using MSS data from Frames 1017-16091 and
1017-16093.
94
-------
UD
0.6 r
ATTENUATION LENGTH OF
DISTILLED WATER
-i30
REFLECTIVITY OF
CHLOROPHYLL
PLANTS
400 500 600 700
COLOR-H V | B G Y 0 R
800 900
NIR
1,000 1,100 1,200
WAVELENGTH (NANOMETERS)
Figure 32. Comparison of the reflectance of chlorophyll-containing plants with the attenuation
length of sunlight in distilled water. From Bressette and Lear (1973). The attenuation
curve is based on Spiess (1970) and the plant reflectivity curve on Katzoff (1962).
-------
The model explains about 83 percent of the variance about the mean
(Table 20). The observed and predicted chlorophyll a_ values, along
with their residuals, are found in Table 21. Although the model is
purported to estimate chlorophyll a, caution must be exercised in
assuming that the model is applicable to other lakes or even to the
same 20 lakes on a different date. The spectral and spatial resolu-
tion of the scanner is low and many factors influence the signal re-
turned to it.
In this case a strong inverse correlation exists between the trophic
indicators LNSECCHI and LNCHLA (-0.012) and some percentage of the re-
turn signal detected by the multispectral scanner is undoubtly due to
the presence of chlorophyll-bearing plants. However, the impact of
different types and concentrations of inorganic suspenoids on the
chlorophyll a^ are not known at this time. The estimates of chlorophyll
a_ derived from the regression model are more properly treated as in-
dex numbers than as absolute values.
96
-------
TABLE 20. ANALYSIS OF VARIANCE TABLE OF A REGRESSION MODEL
FOR THE PREDICTION OF CHLOROPHYLL A LEVELS3
Analysis of Variance
Source
df Sum of Squares Mean Square Calculated F
Total (corrected)
19
23.998
1.263
Regression
Residual
2
17
19.876
4.123
9.938
0.243
40.897
The model was developed using MSS data from Frames 1017-16091 and
1017-16093.
TABLE 21. CHLOROPHYLL A RESIDUALS3
Lake Name
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac la Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
CHLA
(ug/D
29.7
26.5
27.2
0.8
14.1
26.9
4.8
23.0
4.9
1.5
4.1
3.8
13.5
6.9
20.5
5.1
4.1
30.3
29.1
2.1
CHLA
(yg/i )
17.0
29.9
33.7
2.4
26.0
24.4
3.7
41.3
3.5
2.2
3.9
2.1
9.0
8.9
19.0
5.5
5.5
16.9
14.1
2.2
CHLA-Cl^LA
(yg/i )
12.7
- 3.4
- 6.5
- 1.6
-11.9
2.5
1.1
-18.3
1.4
- 0.7
0.2
1.7
4.5
- 2.0
1.5
- 0.4
- 1.4
13.4
15.0
- 0.1
aThe model was developed using MSS data from Frames 1017-16091 and
1017-16093.
97
-------
SECTION VII
LAKE CLASSIFICATION USING LANDSAT-1 MSS DATA
General relationships between MSS data and the trophic state of NES-
sampled lakes, as defined by lake position on the first principal
component axis (PCI value), are examined in this section using
regression analysis, three-dimensional color ratio models, and color-
enhanced photographic products generated through automatic classifica-
tory techniques.
LANDSAT-1 MSS data recorded on 12 different dates (Table 14) were
examined in this investigation. Although the temporal coverage of the
NES-sampled lakes (Table 15) is fragmentary, a relatively coherent time
series exists for 20 lakes in eastern-southeastern Wisconsin. This
section will focus on the MSS frames obtained from the Wisconsin area
on three occasions (9 August 1972, 11 June 1973, and 17 July 1973).
Reference will be made to the other MSS coverage dates and frames which
have been largely relegated to Appendix D.
TROPHIC STATE INDEX PREDICTION USING ERTS-1 MSS DATA
As was discussed in Section IV, trophic state is a multi-dimensional
concept and can not be adequately assessed by measuring any single
indicator. With this in mind, 100 NES-sampled lakes were subjected to
the multivariate technique of principal component analysis using LN
transformed measurements of six trophic state indicators (LNISEC,
LNCHLA, LNTPHOS, LNCOND, LNAAY, LNTON). The position of each lake
on the first principal component axis, the PCI value, is purported to
be a trophic state index. The larger the PCI value, the closer the
water body is to the eutrophic end of the scale and, conversely, the
smaller the PCI value, the closer the lake is to the oligotrophic end
of the scale.
In Section VI the utility of using the LANDSAT-1 MSS data for the pre-
diction of Secchi disc transparency and chlorophyll a_ levels was
examined. While the estimation of specific trophic indicators is of
value, the question arises, "Can the position of a lake on the first
principal component axis be predicted using MSS data?"
Investigations relating to lake appearance as related to trophic status
(Section VI) support the premise that the multitude of interactions
occuring within a lake give the lake volume reflectance an intensity
and spectral curve which is indicative of its trophic state. While it
can be argued that the MSS lacks the spectral resolution to detect some
of the variables incorporated into the trophic state index (such as
98
-------
conductivity and total phosphorus), the elimination of the variables
would make the index less stable and therefore more susceptible to a
large deviation from normal for a given indicator.
PC1-MSS Regression Analyses
The Wisconsin MSS frames are treated in chronological order starting
with those collected on 9 August 1972. Regression models are developed
for the prediction of the trophic state index (PCI) values for 20 NES-
sampled lakes for each sampling date and then for the same lakes on the
basis of mean MSS values for the three dates of LANDSAT-1 coverage.
The product-moment correlation coefficients between the PCI values for
the 20 Wisconsin lakes and the MSS data (colors and color ratios)
extracted from MSS frames recorded on the three MSS sampling dates are
found in Table 22. The correlations of 9 August 1972 and 17 July 1973
are of similar magnitudes.
TABLE 22. CORRELATIONS BETWEEN LANDSAT-1 MSS DATA (COLORS
AND COLOR RATIOS) COLLECTED ON THREE DATES AND
THE TROPHIC STATUS OF 20 WISCONSIN LAKES
Date of LANDSAT-1 Flyover
MSS Colors
and Ratios 9 August 1972 11 June 1973 17 July 1973
GRN
RED
IR1
IR2
GRNRED
GRNIR1
GRNIR2
REDIR1
REDIR2
IR1IR2
0.518
0.722
0.807
0.589
-0.823
-0.806
-0.470
-0.544
-0.026
0.516
0.151
0.533
0.512
0.174
-0.712
-0.540
-0.091
0.084
0.298
0.445
0.479 '
0.721
0.836
0.628
-0.749
-0.820
-0.485
-0.422
0.109
0.479
Numerous regression models were developed during the investigation;
only the "best" models are presented in this report. Criteria used
in the selection of the "best" models included the magnitude of the
coefficient of multiple determination (R2) and the standard error of
estimate. All regression coefficients were required to be significant
at the 0.05 level.
99
-------
The best regression model for the prediction of the trophic state index
(PCI) values of the 20 Wisconsin lakes for 9 August 1972 is:
PCI = 7.682 - 6.074 GRNIR1 + 1.155 GRNIR2
The model explains about 81 percent of the variance about the mean and
has a standard error (s.e.) of estimate of 0.840 (Table 23).
TABLE 23. ANALYSIS OF VARIANCE TABLE OF A REGRESSION MODEL
FOR THE PREDICTION OF THE TROPHIC STATUS OF 20
WISCONSIN LAKES FOUND IN LANDSAT-1 MSS FRAMES
1017-16091 AND 1017-16093 (9 AUGUST 1972)
Analysis of Variance
Source df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
19
2
17
62.555
50.572
11.983
3.292
25.286
0.705
35.867
R2 = 0.8084 x 100
= 80.84% s.e. of estimate = 0.840
When considering the R value of the above model it must be kept in mind
that the PCI was developed using mean values of the ground truth
measurements taken on three occasions during the 1972 open water season.
The MSS data were collected within a few seconds on 9 August 1972.
An examination of the residuals (PCI-PCI) in Table 24 reveals relatively
large absolute values for Middle, Tichigan, Pine, Beaver Dam, and Butte
des Morts. Middle Lake and Butte des Morts are predicted to be in worse
condition (closer to the eutrophic end of the scale) than their PCI
values indicate. The other three lakes are estimated to be in better
condition (closer to the oligotrophic end of the scale) than their PCI
values suggest. The model produced poor results when it was used
to predict the trophic state index values of lakes in other MSS frames.
A regression model was developed for the same 20 Wisconsin lakes using
MSS data collected by the LANDSAT-1 on 11 June 1973. The model:
PCI = 42.761 - 18.423 GRNRED - 18.948 REDIR1 + 3.057 GRNIR2
explains about 70 percent of the variation about the mean and has a
100
-------
TABLE 24. TROPHIC STATE INDEX (PCI) RESIDUALS OF 20 WISCONSIN
LAKES FOUND IN LANDSAT-1 MSS Frames 1017-16091 AND
1017-16093 (9 AUGUST 1972)
Lake Name
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac la Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
PCI
1.68
1.27
2.36
-1.67
3.29
1.48
-1.21
2.45
-1.43
-1.82
-0.62
-0.71
1.27
0.59
3.11
-1.07
-2.29
2.03
1.15
-1.71
PCI
1.41
2.39
2.70
-1.53
2.14
2.13
-0.85
2.87
-0.94
-1.82
-0.89
-1.88
0.58
0.45
1.51
-0.43
-0.56
1.46
1.18
-1.78
PCI -PCI
0.27
-1.12
-0.34
-0.14
1.15
-0.65
-0.36
-0.42
-0.49
0.00
0.27
1.17
0.69
0.14
1 .60
-0.64
-1.73
0.57
-0.03
0.06
standard error of estimate of 1.087 (Table 25). The residuals are
are displayed in Table 26. The model has less practical value for the
prediction of trophic state than the PCI model for 9 August 1972.
The MSS data collected on 17 July 1973 were also used to develop a
regression model for the prediction of the PCI values for the same
lakes with the results:
PCI = 140.430 - 9.375 GRNRED - 98.913 REDIR1 + 45.672 REDIR2 - 57.330
IR1IR2
This model explains about 81% of the variance about the mean and has a
standard error of estimate of 0.892 (Table 27). The residuals are
displayed in Table 28.
Each date of LANDSAT-1 coverage has its unique model for the prediction
of the trophic state of the 20 lakes.
101
-------
TABLE 25. ANALYSIS OF VARIANCE TABLE OF A REGRESSION MODEL
FOR THE PREDICTION OF THE TROPHIC STATUS OF 20
WISCONSIN LAKES FOUND IN LANDSAT-1 MSS Frames
1323-16194 AND 1323-16100 (11 June 1973)
Analysis of Variance
Source df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
19
3
16
62.555
43.649
18.906
3.292
14.550
1.182
12.313
R2 = 0.6978 x 100
= 69.78% s.e. of estimate = 1.087
TABLE 26. TROPHIC STATE INDEX (PCI) RESIDUALS OF 20 WISCONSIN
LAKES FOUND IN LANDSAT-1 MSS Frames 1323-16194 AND
1323-16100 (11 June 1973)
Lake Name
Serial
Number
PCI
PCI
PCI -PCI
Poygan 47
Butte des Morts 48
Winnebago 49
Green 51
Beaver Dam 53
Kegonsa 54
Rock 55
Koshkonong 56
Lac la Belle 57
Oconomowoc 58
Okauchee 59
Pine 60
Nagawicka 61
Pewaukee 62
Tichigan 63
Browns 64
Middle 65
Delavan 66
Como 67
Geneva 68
68
27
36
67
29
47
21
45
43
82
-0.62
-0.71
1.27
0.59
.11
.07
.29
.03
.15
1.19
52
30
-1.71
-0.45
3.29
0.05
-0.63
3.64
-0.11
-0.66
.36
.16
1.10
0.55
1.36
-1.30
-1.22
0.71
1.29
-1.70
-0,
-2,
0.49
-0.25
06
1
-1
.22
0.00
1.42
-0.58
-1
-1
-1
19
32
16
-0.98
1.45
0.17
0.04
1.75
0.23
-1
1
.07
.32
-0.14
-0.01
102
-------
TABLE 27. ANALYSIS OF VARIANCE TABLE OF A REGRESSION MODEL
FOR THE PREDICTION OF THE TROPHIC STATUS OF 20
WISCONSIN LAKES FOUND IN LANDSAT-1 MSS Frames
1359-16091 AND 1359-16094 (17 July 1973)
Analysis of Variance
Source df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
19
4
15
62.555
50.631
11.925
3.292
12.658
0.795
15.922
R2 = 0.8094 x 100
= 80.94% s.e. of estimate = 0.892
TABLE 28. TROPHIC STATE INDEX (PCI) RESIDUALS OF 20 WISCONSIN
LAKES FOUND IN LANDSAT-1 MSS FRAMES 1359-16091 AND
1359-16094 (17 July 1973)
Lake Name
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac la Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
PCI
1.68
1.27
2.36
-1.67
3.29
1.48
-1.21
2.45
-1.43
-1.82
-0.62
-0.71
1.27
0.59
3.11
-1.07
-2.29
2.03
1.15
-1.71
PCI
3.06
1.76
1.73
-1.84
2.15
0.63
-1.20
3.17
-1.48
-1.15
-0.21
-1.59
0.80
0.29
2.83
-0.49
-0.52
0.92
0.70
-1.41
PCI-PCT
-1.38
-0.49
0.63
0.17
1.14
0.85
-0.01
-0.72
0.05
-0.67
-0.41
0.88
0.47
0.30
0.28
-0.58
-1.77
1.11
0.45
-0.30
103
-------
Although the lake color ratios vary from frame to frame (date to date),
it is not unreasonable to expect that over a period of time a lake
would present an average color or appearance, which would be more
representative of the water body's position on a trophic scale. With
this in mind, a new regression model was developed using mean color
ratios derived from the three dates of LANDSAT-1 coverage. For example,
a lake's mean GRNRED color ratio was determined by establishing its
GRNRED ratio for each of the three LANDSAT-1 dates, summing these
ratios and dividing by the number of LANDSAT-1 dates. Stated more
simply,
XGRNRED = (GRNRED1 + GRNRED2 + GRNRED3)/3
where XGRNRED is the mean color ratio for a lake, GRNRED, is the ratio
determined for the first LANDSAT-1 date, GRNRED2 is the ratio for the
second date of LANDSAT-1 coverage, and GRNRED3 is the ratio for the
third date of coverage. The coverage dates are 9 August 1972, 11 June
1973, and 17 July 1973.
The model which best predicts the PCI values for the 20 lakes is:
PCI = 4.127 - 6.623 XGRNRED - 3.511 XREDIR2 + 8.040 XIR1IR2
It explains about 80 percent of the variation about the mean and has a
standard error of estimate of 0.887 (Table 29). The residuals are
displayed in Table 30. Relatively large residuals (absolute values)
exist for Middle, Pine, and Butte des Morts. The elimination of the
11 June 1973 data and use of average ratios developed from the two
remaining dates (9 August 1972 and 17 July 1973) would result in a
better model.
Regression models were also developed for NES-sampled lakes in Minnesota
and New York where LANDSAT-1 MSS coverage was available. The best
model, as measured by the criteria stated previously, for each date of
LANDSAT-1 MSS coverage is displayed in Table 31 along with the models
for the Wisconsin lakes.
Although a model is included for all of the LANDSAT-1 MSS coverage
dates, except for the two dates with very small sample sizes, some of
the models are clearly inadequate. For example, the models with an
R2 less than about 0.80 (80 percent) do not have much predictive value.
This effectively eliminates a model for the Wisconsin lakes (11 June
1973) and several models for the Minnesota lakes (6 October 1972, 28
May 1973, 3 July 1973). Another model, 19 August 1972 (New York) is of
dubious value because the number of observations is small (N = 7), and
there are only four degrees of freedom associated with the residuals.
104
-------
TABLE 29. ANALYSIS OF VARIANCE TABLE OF A REGRESSION MODEL FOR THE
PREDICTION OF THE TROPHIC STATUS OF 20 WISCONSIN LAKES
USING MEAN MSS COLOR RATIOS FROM THREE DATES (9 AUGUST
1972, 11 JUNE 1973, 17 JULY 1973)
Analysis of Variance
Source df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
19
3
16
62.555
50.003
12.529
3.292
16.676
0.783
22.575
R2 = 0.7997 x 100 = 79.97% s.e. of estimate = 0.887
TABLE 30. TROPHIC STATE INDEX (PCI) RESIDUALS OF 20 WISCONSIN
LAKES FROM A REGRESSION MODEL INCORPORATING MEAN MSS
COLOR RATIOS FROM THREE DATES (9 AUGUST 1972, 11 JUNE
1973, 17 JULY 1973)
Lake Name
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac la Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
PCI
1.68
1.27
2.36
-1.67
3.29
1.48
-1.21
2.45
-1.43
-1.82
-0.62
-0.71
1.27
0.59
3.11
-1.09
-2.29
2.03
1.15
-1.71
PCI
1.66
2.45
1.77
-1.61
3.05
1.33
-0.93
3.32
-1.05
-1.23
-0.47
-1 .99
0.42
0.29
1.48
-0.69
-0.36
1.45
0.98
-1.74
PCI -PCI
0.02
-1.18
0.59
-0.06
0.24
0.15
-0.28
-0.87
-0.38
-0.59
-0.15
1.28
0.85
0.30
1.63
-0.40
-1.93
0.58
0.17
0.03
105
-------
TABLE 31. REGRESSION MODELS DEVELOPED FOR THE ESTIMATION OF TROPHIC STATE USING LANDSAT-1 MSS DATAe
Date, Frame(s) ••• Number
and Area of Lakes
Regres'sion Model
1359-16091
1359-16094
Wisconsin
1017-16091
1017-16093
1323-16194
1323-16100
1359-16091
1359-16094
14 August 1972
1022-16373
Minnesota
-, Standard
R x 100 Error of Appendix
Estimate
9 August 1972
1017-16091
1017-16093
Wisconsin
11 June 1973
1323-16194
1323-16100
Wisconsin
17 July 1973
/\
20 PCI
/\
20 PCI
= 7.682 - 6.074 GRNIRl + 1.155 GRNIR2
= 42.761 - 18.423
GRNRED - 18.948 REDIR1
80.84% 0.840
69.78% 1.087
Dl
D3
+ 3.057 GRNIR1
/\
20 PCI
= 140.430 - 9.375
GRNRED - 98.913 REDIR1
80.94% 0.892
D4
+ 45.672 REDIR2 - 57.330 IR1IR2
20 PCI = 4. 127 - 6.623 XGRNRED - 3.511 XREDIR2 79.97% 0.887
+ 8.040 XIR1IR2
/\
11 PCI = 13. 150 - 10.626 GRNIR1 + 2.327 GRNIR2 96.06% 0.456 D5
-------
TABLE 31. REGRESSION MODELS FOR THE ESTIMATION OF TROPHIC STATE USING LANDSAT-1 MSS DATA (continued)
Date, Frame(s) Number
and Area of Lakes
6 October 1972 12
1075-16321
Minnesota
8 October 1972 7
1077-16431
Minnesota
28 May 1973 11
1309-16325
Minnesota
3 July 1973 12
1345-16322
Minnesota
4 July 1973 7
1346-16381
Minnesota
19 August 1972 7
1027-15233
New York
Regression Model
PCI = 11.533 - 7. 132 REDIR1
PCI = 35.509 - 18.548 GRNRED
/\
PCI = -16.537 + 9.844 IR1IR2
/\
PCI = 10.544 - 7.240 REDIR1
/\
PCI = 11.715 - 8.277 REDIR1
/\
PCI = -4.891 - 8.805 GRNIRl
+ 19-301 REDIR1
~ Standard
R x 100 Error of
Estimate
43.91% 1.540
95.12% 0.247
49.33% 1.525
70.09% 1.117
92.34% 0.669
82.82% 0.740
Appendix
D6
D7
D8
D9
D10
Dll
aRegression models were not constructed for Frames 1036-16152 (28 August 1972, Wisconsin) and 1080-
15180 (11 October 1972, New York) due to the relatively small number of observations.
-------
An examination of the remaining models indicates that they all differ
from one another. Attempts to predict the PCI values for a particular
LANDSAT-1 coverage date, using a model from another date, produced
statistically insignificant results.
Although their regression coefficients are different, some models have
the same variables in common. For example, the 9 August 1972 (Wis-
consin) model and the 14 August 1972 (Minnesota) model incorporate the
variables GRNIR1 and GRNIR2. Both models have an R2 exceeding 0.80 and
are of value in predicting the PCI values of water bodies contained in
their respective frames.
The models for 3 July 1973 (Minnesota) and 4 July 1973 (Minnesota)
contain a single independent variable, IR1. Yet, the first model has
an R2 of about 70 percent and the latter model an R2 of about 92 per-
cent. Some of the lakes are common to both models. An attempt to
use the 4 July 1973 model to predict the PCI values of the NES lakes in
the 3 July 1973 frame drew statistically insignificant results.
The degree of success in the prediction of lake trophic state, as
defined by a lake's position on the first principal component axis,
varies considerably from date to date. Several factors could account
for the variability.
Lakes are, by their very nature, dynamic and can change significantly
in appearance over a matter of days or weeks. Algal blooms, the
growth of aquatic macrophytes, the influx of silt carried by rain-
swollen streams, and, in the case of shallow lakes with sufficient
fetch, sediments churned up by wind-induced turbulence can produce
changes in the volume reflectance detectable by both human observers
and other sensors such as the MSS. Super-imposed on the variations
associated with lake dynamics are variations related to atmospheric
conditions and solar angle.
The impact of antecedent precipitation on lake volume reflectance can
not be ascertained with any degree of certainty in this study. The
relatively poor regression model for the 3 July 1973 (Minnesota) may
be a consequence of the heavy rains on 2 July 1973, but this is largely
conjecture.
THREE-DIMENSIONAL MSS COLOR RATIO MODELS
In the preceding section, several regression models were developed to
predict the trophic status of selected lakes using MSS ratios. The
assumptions were made that the PCI values adequately represent the
position of the lakes on a trophic scale and that lake phenomena that
correlate with the index are detectable using the MSS.
108
-------
A less sophisticated but practical approach to evaluating relation-
ships between MSS data and trophic state involves the visual examina-
tion of MSS data patterns in light of a general knowledge of the lakes
as well as their trophic state index values (PCI). Although this
could be done through the use of data matrices, a graphic approach is
favored because it is very conducive to pattern detection and inter-
pretation. An examination of the various MSS ratios incorporated into
the regression models in Table 31 indicates that the MSS ratios GRNIR1,
GRNRED, and REDIR1 might be used to advantage.
The three-dimensional models in this section were produced using the
same program and equipment as in the PCA ordination model found in
Section IV. The numerals inside the "ball" are the lake's serial
number and those near the lake's name represent its PCI value. Lakes
with a serial number greater than 100 fell outside the scope of the
PCA ordination and therefore lack a PCI value. Although a MSS color
ratio model was developed for each date of LANDSAT-1 coverage, the
Wisconsin lakes will serve as a focal point. The other models are
found in Appendix D.
The color ratio model for 9 August 1972 (Wisconsin: 1017-16091,
1017-16093) is found in Figure 33. The model can be examined using
both a general knowledge acquired about the lakes and their PCI
values as guides. There is a very definite trend for the color ratios
to increase as one moves from lakes considered to be located near the
eutrophic end of the scale (e_.£., Beaver Dam) toward those situated
more closely to the oligotrophic end (e_.c[., Green, 51).
It is unrealistic to expect complete agreement between the position
of the lakes in the color ratio model and their respective PCI values.
In addition to the problems created by the dynamic nature of the
lakes, some additional uncertainty is generated by the sampling
methods. The lake MSS data for the three-dimension model were
acquired by sensing the lake body, at least to Secchi depth (or to the
bottom if the Secchi depth is greater than the water depth), on a
pixel-by-pixel basis. The PCI values were derived from ground truth
collected at selected points (stations) ranging in number from one
(e_.£o, Middle Lake) to nine (Lake Winnebago). As will be seen in the
next section, the number of sampling sites and their location on a
lake can have a significant effect on the lake's trophic index value.
In addition, the sensor data represent lake phenomena at a single
point in time; the PCI values were derived using annual mean values.
Assuming that the PCI value of Middle Lake is representative of its
trophic state, it is "out of position" relative to the other lakes in
the 9 August 1972 model. The lake's color ratio coordinates are
indicative of a lake situated more closely to the eutrophic end of the
trophic scale. Several factors may be responsible for this apparent
misclassification.
109
-------
2.45
Koshkonong
-I . 43
Lac La Belle
I 27
Butte Des Mortsf^)
3.29
Tichigan L Beaver Dam
-2.29
1.27
. Nagaw icka
-0 .62
-I .82
Oconomo woe -j>
- I .21
Rock
Pine
-0.71
-1.71
Geneva
Green (
G R N : I R 1
Figure 33. Three-dimensional color ratio model for 9 August 1972. The 20 Wisconsin lakes were
extracted from LANDSAT-1 MSS Frames 1017-16093 and 1017-16091. The frames are in
juxtaposition on the same flight line.
-------
As discussed previously, the trophic scale does not directly incorpor-
ate the extent of aquatic macrophytes and algal organisms in the lakes,
nor does it include a direct measure of lake morphometry. The sensor
may very well be "seeing" large masses of plants and/or the lake
bottom. Middle Lake is known to have weed problems and it has exten-
sive shallows. The incorporation of some direct measure of aquatic
weeds into the trophic state index would shift Middle Lake toward the
eutrophic end of the scale and bring the lake index value in closer
agreement with the lake's coordinates in the color ratio model.
The 11 June 1973 model of the same 20 lakes is shown in Figure 34 along
with three other lakes (Mendota, Monona, and Waubesa). Many of the
lakes have shifted their position significantly (e_.g_., Winnebago,
Kegonsa, Geneva, Rock, and Oconomowoc). The color ratio-trophic state
relationships so evident in the three-dimensional model of 9 August
1972 are not as obvious in 11 June 1973. Efforts to develop a
regression model, using the PCI values as the dependent variable in the
preceding section, yielded statistically insignificant results.
The Wisconsin lakes are incorporated into a three-dimensional color
ratio model using MSS data from 17 July 1973 (Figure 35). Although
not identical in all respects, the model bears a marked resemblence
to the 9 August 1972 model. The positional change of Lake Poygan may
be due in part to the large portion of cloud coverage (approximately
50 percent).
The model in Figure 36 represents the mean color ratio relationships
among the 20 Wisconsin lakes using the MSS data for three dates (9
August 1972, 11 June 1973, 17 July 1973). The mean ratios were
determined in the previous section. The model may be thought of
as a representation of the general appearance of the lakes. A more
extended time series is desirable, but cloud cover and poor quality
MSS bands have made this impossible despite the 18-day repetitive
coverage cycle.
A three-dimensional color ratio model was constructed for each of the
remaining dates of LANDSAT-1 MSS coverage (Table 15). The models are
found in Appendix D. It will be noted in some of the models of
Minnesota lakes (14 August 1972, 6 October 1972, 8 October 1972) that
certain lakes (Wagonga, Silver) are isolated from the other lakes.
These lakes have IR1 DN values which exceed their respective RED DN
values. The lakes are at the extreme end of the trophic scale and
are sometimes referred to as being hypereutrophic.
An examination of both the three-dimensional color ratio models and
their associated regression models suggests that the utility of the
MSS for the estimation of trophic state is dependent to a substantial
degree upon the time of year. The Wisconsin and Minnesota frames
recorded relatively early during the open water season (28 May 1973,
111
-------
ro
G R N : I R 1
Figure 34. Three-dimensional color ratio model for 11 June 1973. The 23 Wisconsin lakes were
extracted from LANDSAT-1 MSS Frame 1323-16100. Three of the lakes (Mendota, Monona,
and Waubesa) fall outside the scope of the investigation, but are included because
they are well-known by lake scientists.
-------
3-29
Beaverdar" (53
.27
Nagawicka
1.4 8
Kegonsa/
G R N • I R 1
Figure 35. Three-dimensional color ratio model for 17 July 1973. The 20 Wisconsin lakes were
extracted from LANDSAT-1 MSS Frames 1359-16091 and 1359-16094.
-------
©'
2.45
^Koshkonong
©
1.15
Como
. 68
•sPoygan
2.36
Winnebago
-1.43
^ac La Belle
-2 9
3.11
T i c h i g a n
' -2 7
i • t 3 ,
Beaver Dam ^~,Butte DesMorts
1 (<«1
1.27
Nagawicka^j
2.03
De lavan
0.59
77)Pewaukee
-1.07
-2.29
.Middle
— 0.62
^Okauche e
-1.21
_Rock
-1.82 „
OconomowocQ
-1.71
. Genev a
-0.71
-1.67
Greenfs,
CRN IR1
Figure 36. Three-dimensional color ratio model for three dates of LANDSAT-1 MSS coverage (9 August
1972, 11 June 1973, 17 July 1973). The model was developed using mean color ratios
derived from MSS data collected on the three dates.
-------
11 June 1973) have spectral curves which correlate poorly with the
lake PCI values. The correlations are much stronger in the case of
MSS frames recorded later in the season (9 August 1972, 14 August
1972, 17 July 1973, 8 October 1972) when the contrast between lakes
at different points on the trophic scale tend to be at a maximum.
LAKE CLASSIFICATION USING MSS DATA IN CONJUNCTION
WITH AN AUTOMATIC DATA PROCESSING TECHNIQUE
Up to this point the lake MSS data have been examined using regression
techniques and three-dimensional color ratio models. The investiga-
tions were undertaken using mean color values and mean color ratios
for each LANDSAT-1 date. The lakes were treated as entities without
regard to differences which might be present within them. Automatic
data processing techniques are well-suited, not only for classifying
lakes, but also for classifying different types of water within
individual lakes (Boland, 1975; Boland and Blackwell, 1975).
It is becoming increasingly common to use automatic data processing
(ADP) techniques in remote sensing studies (e_.g_., Hoffer, et al.,
1972). ADP techniques merit consideration because they permit the
reduction of large masses of data in realistic time periods and add
objectivity to the classificatory process.
With the advantages of ADP in mind, an effort was undertaken by
Blackwell (1974, personal communication) to apply the techniques to
the 20 lakes found in MSS Frames 1017-16091 and 1017-16093 (9 August
1972). The equipment and software at the JPL/IPL were used to
process the MSS data reported here and produce hard copy in the form
of black and white photographs and color-enhanced prints. The method-
ology employed is briefly described below.
Utilizing a LARS-developed (Laboratory for Applications of Remote
Sensing, Purdue University) spectral pattern-recognition-algorithm,
the IPL IBM 360/44 was trained using the GRN, RED, and IR1 MSS data
in conjunction with the lake PCI values. Preliminary processing
indicated that the IR2 band data would be of little value in distin-
guishing one lake from another or one area of a lake from another
area of the same water body. The IR2 data were not included, thereby
reducing the amount of CPU time required for classification.
The number of spectral classes was set by establishing one class for
each different PCI value among the 20 Wisconsin lakes. This resulted
in the formation of 19 different classes (Table 32). Butte des Morts
and Nagawicka have the same PCI value and were assigned to the same
class. The computer was statistically trained to recognize each lake
as belonging to a particular class. For example, the computer was
trained to perceive Beaver Dam Lake as belonging to Class 1, Tichigan
as Class 2, ..., Middle Lake as Class 19. Each pixel in the 20 lakes
115
-------
was then classified by the computer into one of the 19 classes,
results of the classification procedure are found in Table 33.
The
TABLE 32. LAKE TROPHIC STATE INDEX CLASS ASSIGNMENTS
FOR THE ADP TECHNIQUE
Lake Name
Beaver Dam
Tichigan
Koshkonong
Winnebago
Delavan
Poygan
Kegonsa
Butte des Morts
Nagawicka
Como
Pewaukee
Okauchee
Pine
Browns
Rock
Lac la Belle
Green
Geneva
Oconomowoc
Middle
Serial
Number
53
63
56
49
66
47
54
48
61
67
62
59
60
64
55
57
51
68
58
65
Computer Trained
PCI to Recognize as
Value Class:
3.29
3.11
2.45
2.36
2.03
1.68
1.48
1.27
1.27
1.15
0.59
-0.62
-0.71
-1.07
-1.21
-1.43
-1.67
-1.71
-1.82
-2.29
1
2
3
4
5
6
7
8
8
9
10
11
12
13
14
15
16
17
18
19
All of the pixels in a homogeneous lake would be classified as falling
into the class for which the lake served as a training area. It is
very unlikely to find a lake that has the same trophic characteristics
throughout its areal extent. Some indication of the heterogeneous
nature of the 20 Wisconsin lakes may be obtained by examining Table 33.
Beaver Dam Lake, for example, has 69.4 percent of its pixels classified
as belonging to Class 1, 10 percent to Class 2, ..., and 1.3 percent to
Class 7. Lake Kegonsa exhibits the least heterogeneity with 87.4 per-
cent of its pixels falling in Class 7. The percentages expressed here
should be treated as approximations because the "sixth line" banding
affects the results of any such classification scheme. If the classi-
ficatory results had indicated homogeneous conditions in each lake,
116
-------
TABLE 33. ADP RESULTS FOR 9 AUGUST 1972 USING A 19 CLASS CLASSIFICATION
Lake Name
Beaver Dam
Tichigan
Koshkonong
Winnebago
Delavan
Poygan
Kegonsa
Butte des Morts
Nagawicka
Como
Pewaukee
Oka lichee
Pine
Serial
Number
53
63
56
49
66
47
54
48
61
67
62
59
60
PCI Trained
Value as Class:
3.29
3.11
2.45
2.36
2.03
1.68
1.48
1.27
1.27
1.15
0.59
-0.62
-0.71
1
2
3
4
5
6
7
8
8
9
10
11
12
Six Major
1
(69.4)
2
(41.6)
7
(44.5)
4
(18.0)
5
(66.0)
7
(24.0)
7
(87.4)
2
(29.3)
15
(41.3)
7
(39.1)
10
(35.0)
11
(32.7)
12
(35.9)
10
(13.0)
5
(17.0)
17
(12.6)
10
(16.3)
3
(8.0)
6
(18.3)
9
(7.4)
8
(23.3)
10
(16.5)
9
(31.2)
14
(14.9)
16
(19.5)
16
(30.9)
Classes Found in the Lake
(Percentage)
2
(9.8)
1
(16.3)
16
(6.0)
5
(13.8)
8
(6.1)
9
(13.4)
6
(3.0)
5
(18.4)
5
(9.3)
1
(7.5)
1
(13.0)
14
(13.0)
11
(4.6)
19
(3.2)
4
(8.5)
12
(5.1)
2
(9.8)
7
(4.2)
8
(10.1)
3
(1.2)
9
(7.8)
4
(8.3)
2
(5.0)
11
(8.9)
10
(11.8)
17
(4.1)
4
(2.5)
8
(8.5)
2
(4.8)
7
(9.5)
1
(4.0)
5
(8.0)
2
(0.1)
3
(7.0)
17
(6.0)
4
(3.4)
13
(5.2)
1
(6.6)
19
(4.1)
7
(1.3)
7
(3.9)
3
(4.7)
1
(9.1)
2
(3.5)
4
(6.7)
1
(0.1)
6
(5.8)
1
(5.0)
8
(3.4)
19
(5.2)
12
(5.8)
13
(3.7)
-------
TABLE 33. ADP RESULTS FOR 9 AUGUST 1972 USING A 19 CLASS CLASSIFICATION9 (continued)
OD
Lake Name
Browns
Rock
Lac la Belle
Green
Geneva
Oconomowoc
Middle
Serial
Number
64
55
57
51
68
58
65
PCI
Value
-1
-1
-1
-1
-1
-1
-2
.07
.21
.43
.67
.61
.82
.29
Trained
as Class:
13
14
15
16
17
18
19
Six Major
11
(22.0)
16
(31.1)
15
(43.0)
16
(45.0)
17
(38.0)
16
(23.6)
12
(22.2)
13
(17.1)
12
(15.6)
10
(11.9)
12
(27.8)
16
(18.3)
12
(19.2)
19
(15.4)
Classes Found in the Lake
(Percentage)
19
(15.5)
14
(12.7)
14
(11.8)
14
(7.7)
12
(15.6)
14
(12.4)
13
(12.8)
12
(10.0)
11
(12.5)
7
(5.3)
17
(5.7)
14
(12.7)
17
(9.4)
1
(11.1)
16
(6.5)
15
(9.9)
5
(4.9)
11
(5.2)
18
(5.4)
15
(8.7)
11
(10.2)
1
(6.1)
17
(4.6)
11
(4.8)
2
(1.9)
10
(2.3)
11
(7.9)
16
(10.2)
The six trophic classes listed for each lake were assigned most of the pixels representing the
lake. The percentage of pixels assigned to each class is given in parentheses.
-------
the analysis could stop at this point. However, this is not the
case, and the question arises: "What trophic-related patterns exist
in each lake?" This necessitates the development of some sort of
imagery.
Images of the machine-classified lakes can be produced in the form of
line printer copy using different symbols to represent the various
trophic classes and also as photographic prints and transparencies.
Output in the form of photographs was selected because they are
compact, easily handled, have much greater resolution than line
printer copy of equal size, and are readily interpretable, particularly
when in color.
The ADP-classified lakes (9 August 1972) are displayed in Figure 37
using 19 gray-levels, one for each class. Class 1 is located toward
the eutrophic end of the trophic scale (black) and Class 19 (white)
is toward the oligotrophic end. It would be incorrect to call a
Class 19 pixel or lake oligotrophic because none of the lakes examined
are considered to possess the necessary attributes. The pronounced
linear features are artifacts introduced by a defect in the MSS and
are generally referred to as "sixth line" banding. It is important
to overlook their presence when studying the patterns present in the
lake images.
Figure 38 is a color-enhanced version of the 19-class classification
of the 20 lakes (9 August 1972). Black has been assigned to represent
Class 1, and each of the remaining 18 classes has been assigned its
unique color. The colors were produced by using different ratios of
the three primary colors: blue, green, and red. Detailed information
regarding the principles involved is found in the work by Committee
on Colorimetry (1966).
Differences among and within the lakes are readily apparent. Some of
the lakes (Kegonsa and Beaver Dam) present a relatively homogeneous
appearance. Others (Winnebago and Poygan) exhibit a diversity of
trophic classes. Some of the lakes have features which bear
mentioning.
The appendix-like structure which appears attached to the northeast
quadrant of Delavan Lake is the entry point of Jackson Creek, the
major stream feeding the lake. Its waters, known to be nutrient-rich
through contributions from sewage treatment plants and agricultural
drainage, have been placed in Class 1.
In this study, Lake Tichigan has been defined to include the lake
proper and what is commonly referred to as the "widening" in the Fox
River. The lake proper has been assigned to Class 5 and the "widening"
to several classes including 1, 2, and 4. Ground truth measurements
indicate that the "widening" has a lower Secchi disc transparency and
a substantially high chlorophyll a_ level.
119
-------
EPA JPL LAKI
EPT: nss FRAME 101? - it.o9i-1909:-:
TE:-:TAD - ITPETCH
GREY LEVEL CLASSIFICATION
Y P R D G R A M
09 AUG •:! WISCONSIN >
RJB SEP 11, 1974 081941 JPL IPL
Figure 37. ADP-classified lakes (9 August 1972), using 19 gray-levels,
one for each class.
120
-------
r=-"E
- = r-
••* •»» »
. r - - r PRQ
- I-: -i : -: "-:
Figure 38. A color-enhanced version of the 19-class classification of
the 20 lakes (9 August 1972).
121
-------
The Class 1 water found along the northern shore of Lake Pewaukee may
be related to the presence of algae and rooted emergent plants. The
helicopter-borne survey teams reported algal scum covering the surface
of the northern portion of the lake on 21 June 1972 and 19 August 1972.
Heavy growths of emergents covered the lake shallows.
The appendix-like portion of Green Lake located at its northeast end is
the area into which Silver Creek flows. The area receives a substantial
nutrient load from a sewage treatment plant and the surrounding agri-
cultural lands. Its pixels have been classified as belonging to Class
1 and Class 2.
White areas within the lake images are indicative of either clouds or
land-related phenomena. The white area in the northeastern portion of
Lake Winnebago represents cloud cover. The north-south linear feature
located in the eastern end of Lake Butte des Morts is a causeway.
Complete accord does not exist between the trophic index values of the
20 lakes and the results of the ADP technique. The disparity is very
evident in the cases of lakes Nagawicka, Koshkonong, and Oconomowoc
where few, if any, pixels were found that fell into the class for which
the lake served as a training area. Middle Lake contains pixels
classified as belonging to Class 19, but they constitute only 15.4 per-
cent of the total. This is not surprising because there is an indica-
tion in the three-dimensional color ratio model and the PCI regression
model for 9 August 1972 that a disparity exists between some of the
lake PCI values and their MSS data. The use of a smaller number of
classes may very well yield more consistent results. This is an area
in need of additional study.
The color-enhanced imagery, produced through an ADP technique, should
prove to be of value, not only in comparing lakes with each other, but
also for supplying information which can be used in the selection of
future lake sampling sites. Further refinement should make the imagery
a valued tool in lake survey and monitoring activities.
122
-------
SECTION VIII
GENERAL SUMMARY
This research has focused on relationships between Earth Resources
Technology Satellite-One (LANDSAT-1) multispectral scanner (MSS) data
and the trophic status of selected lakes in Minnesota, Wisconsin,
Michigan, and New York sampled by the U.S. Environmental Protection
Agency's National Eutrophication Survey. Analyses were directed
toward lake classification on the basis of ground truth, and the pre-
diction of lacustrine trophic state indicator magnitudes and trophic
state using MSS data.
Initially, 100 NES-sampled lakes were analyzed to ascertain their
trophic status. Trophic indicators selected as parameters in the
analyses included: chlorophyll a_, conductivity, total phosphorus, total
organic nitrogen, the inverse of Secchi disc transparency, and the yield
of a standard algal assay procedure. Natural logarithm transformations
were made on the data to produce a more-nearly normal distribution.
A complete linkage algorithm (McKeon, 1967) was used in conjunction
with squared-Euclidian distance to examine the trophic status of the
lake population for the presence of natural clusters or groups. The
results are displayed in the form of a dendrogram (Figure 15).
The lake trophic indicator data were also subjected to a principal
components analysis to reduce the dimensionality of the data from six-
dimensional hyperspace to space of three or fewer dimensions. The
eigenvectors and their associated eigenvalues were extracted from a
p x p Pearson product-moment correlation matrix. The first principal
component (normalized eigenvector) accounts for approximately 68 percent
of the variation in the data. The second and third components represent
about 14 percent and 8 percent of the variation, respectively. The
results of the principal components analysis were used to ordinate the
lakes in one-, two-, and three-dimensional space. The three-dimensional
ordination is displayed as a "ball and wire" model (Figure 17).
A multivariate trophic state index was developed by evaluating the
first principal component for each of the 100 NES-sampled lakes. The
resultant values (PCI) are indicative of each lake's respective position
on a multivariate trophic scale (the first principal axis). The larger
the PCI values, the closer the lake lies toward the eutrophic end of the
scale. The coefficients of the first component are nearly equal in
magnitude, suggesting that the first principal component represents a
general measure of trophic state.
123
-------
LANDSAT-1 MSS data, collected on 12 different dates from Minnesota,
Wisconsin, and New York, were extracted from computer-compatible
magnetic digital tapes (CCT's) using VICAR, a sophisticated software
system at the Image Processing Laboratory (IPL), a support facility
of the Jet Propulsion Laboratory (JPL). Each lake image was extracted
from its terrain matrix through the use of a binary mask technique
which employs a digital number (DN) level of 28 (8-bits of precision;
256 DN levels) as the water-land cutoff point. Descriptive statistics
were computed for each lake including total number of pixels, mean DN
values, standard deviations, and histograms. This was accomplished for
each of the four MSS bands (GRN, RED, IR1, and IR2). A photographic
concatenation was produced of the lakes extracted from each frame.
Regression models were developed to predict the magnitude of two trophic
state indicators (Secchi disc transparency and chlorophyll a_) using MSS
data from two dates, 9 August 1972 (Wisconsin) and 14 August 1972
(Minnesota). A regression model was developed for each date of
LANDSAT-1 MSS coverage using the multivariate trophic state index (PCI)
as the dependent variable and the MSS color ratios as independent
variables. Most of the modelling effort was directed toward NES-sampled
lakes in eastern-southeastern Wisconsin. The Wisconsin area was selec-
ted because good LANDSAT-1 MSS coverage was available for a group of
20 lakes on three different dates.
A more flexible approach toward the study of PCI-MSS relationships was
undertaken using three-dimensional MSS color ratio models. The MSS
ratios GRNIR1, GRNRED, and REDIR1 were used to develop the three-
dimensional "ball and wire" models. A model was constructed for each
date of MSS coverage. In addition, a model was created for the
Wisconsin lakes using mean color ratio values for the three sampling
dates (9 August 1972, 11 June 1973, and 17 July 1973; Figure 36).
Automatic data processing (ADP) techniques were used in conjunction
with the PCI values for 20 Wisconsin lakes and MSS data 9 August 1972)
to classify the lakes. A computer was programmed to recognize 19
trophic classes using the MSS color information contained in the 20
lakes. The classification was accomplished on a pixel-by-pixel basis
using the MSS GRN, RED, and IR1 data. Output is in the form of
statistical data and photographic products including 19-step gray
scale prints (Figure 37) and 19-class color-enhanced photographs
(Figure 38).
LAKE CLASSIFICATION USING GROUND TRUTH
The examination and classification of large numbers of lakes within
realistic periods of time necessitates the use of automatic data
processing techniques which incorporate classificatory methods, such
as cluster analysis and principal components ordination.
124
-------
The complete linkage algorithm used in this research did not produce
three well-isolated clusters (groups) that can be readily associated
with the traditional trophic classes of lakes - oligotrophic, mesotro-
phic, and eutrophic. This apparent failure is due in part to the
lakes used in the study. Although the 100 lakes were selected to
cover the range of lakes sampled during 1972 by the National Eutrophi-
cation Survey, the NES-sampled lake population is heavily weighted
toward lakes with relatively poor water quality. The three-dimensional
principal component ordination does not contain large well-defined
clusters. The ordination results support the findings of the cluster
analysis.
It is likely that there are no sharp discontinuities which clearly
separate the members of a large population of lakes into the three
commonly-referred-to trophic classes. Lakes of such a population
may be characterized as constituting a hyper-dimensional cloud
possessing a low degree of organization.
Cluster analysis techniques are a "heavy handed" approach to the lake
classification problem because they will lead to clusters even if the
data are random. The principal component analysis ordination does not
presume the existence of clusters. The two methods are complementary
and should be used in conjunction with one another.
The concept of a multivariate trophic state index, as defined by the
position of a lake on the first principal component axis, merits
further consideration. However, it must be kept in mind that a less
sophisticated method, the mean composite rank (MCR) index, yielded
similar results as measured by lake rank. The MCR system has the
advantages of conceptual simplicity and ease of computation.
MSS Estimation of Lake Area and Selected Trophic State Indicators
The LANDSAT-1 MSS is an effective tool for the determination of the
number and the areal extent of lakes. The use of a DN level of 28
(8-bits of precision) as the water-land cutoff point for the generation
of the binary mask in conjunction with a pixel conversion factor of
0.48 (1 pixel = 0.48 hectares) gives lake area estimates generally
within 10 percent of values derived from topographic sheets.
While the use of near-concurrent ground truth precludes more precise
estimates, it has been demonstrated that good predictions of Secchi
disc transparency can be achieved through the incorporation of MSS
color ratios into regression models.
A measure of the chlorophyll a_ level can be obtained by using MSS color
ratios as independent variables in regression models. However, caution
125
-------
must be used in the interpretation of the chlorophyll model predic-
tions. The predicted values should be treated as a crude index
rather than accurate estimations of chlorophyll a_.
Models for the prediction of Secchi disc transparency and chlorophyll
a_ are unique to the date of LANDSAT-1 MSS coverage. The models do not
give good predictions when used with MSS data collected on different
dates. Ground truth must be collected concurrently or near-
concurrently with MSS coverage, and then the appropriate models must
be constructed.
Although the need for ground truth places a restriction on the use of
MSS data for the prediction of trophic state indicators, it still has
utility in regions where lake concentrations result in the inclusion
of hundreds and even thousands of lakes within a single MSS frame.
Under such circumstances the collection of ground truth from a small
number of "bench mark" lakes for the development of regression models
would pay handsome dividends.
MSS Prediction of Lacustrine Trophic State
The MSS data can be used to give fair to very good estimates of
lacustrine trophic state as defined by lake position on the first
principal component axis. The regression models, developed using MSS
color ratios as independent variables, are unique to the date of
LANDSAT-1 coverage. Predictions made employing a specific model in
conjunction with MSS data from other dates were statistically in-
significant. Models developed from MSS data collected early in
the open-water season are inadequate; better models were constructed
using MSS data collected later in the season when the contrasts between
lakes at different points on the trophic scale tend to be maximized.
Extraction Techniques and Products
While it is possible to extract MSS data related to lake trophic state
and trophic indicators from EROS-supplied imagery, efforts to use this
imagery are seriously impeded because the photographic products lack
the requisite radiometric fidelity, a consequence of the scale compres-
sion induced by condensing the MSS intensity resolution from 127 levels
(7-bit precision) to 16 gray levels. Data extracted from computer-
compatible tapes (CCT's) must be used if the maximum benefits are to be
derived from the LANDSAT-1 MSS. This is particularly important in
water quality-related studies because the full range of water quality
differences is contained in a relatively small number of DN levels.
The computer-compatible tape approach to the problem of MSS data
extraction has the added advantage of eliminating the errors introduced
by differences in photographic products and the microdensitometer. CCT
data extraction and processing techniques, which produce both MSS data
126
-------
statistics and photographic products, greatly increase the utility of
the multispectral scanner.
Photographic products permit the rapid visual assessment of lake
coverage by providing greater spatial resolution and a fraction of the
bulk characteristic of line printer copy. It has been demonstrated
that automatic digital processing techniques can produce trophic
classifications which, when displayed as photographic imagery, can be
incorporated into studies of lake water quality, water circulation
patterns, and supply information for the selection of lake water
sampling sites. The advantages of using CCT-produced photographs are
particularly evident when the products are in the form of color-enhanced
prints and transparencies.
LANDSAT-1 Coverage and Quality
Although LANDSAT-1 gives 18-day repetitive coverage, good frames were
available for only a few days in the study area encompassing Minnesota,
Wisconsin, Michigan, and New York. On many occasions, excessive cloud
cover (greater than 10 percent) was present or one or more of the MSS
bands was rated as poor or missing. Good coverage was virtually non-
existent for NES-sampled lakes in Michigan during 1972.
While it is recognized that the number of days with excessive cloud
cover is a function of a region's weather and climate, and therefore
its geographic location, - some regions (southwestern U.S.) have viewing
conditions which are consistently better than other regions (north-
eastern U.S.) - it is apparent that good LANDSAT-1 MSS coverage is
available on a very erratic basis, making systematic time-series studies
difficult if not impossible in the study area and, indeed, for much of
the eastern United States.
Information relating to water quality is found in each of the four MSS
bands (GRN, RED, IR1, and IR2). However, the DN levels sensed for
water bodies are few in number and are located at the lower end of the
intensity scale. Surface water resource investigations would derive
substantial benefits if the instrument gain were to be increased on the
MSS. It is possible to switch to high gain in the GRN and RED bands.
127
-------
SECTION IX
REFERENCES
1. Anderberg, M. R. 1973. Cluster analysis for applications. New
York: Academic Press. 373 pp.
2. Anderson, T. W. 1958. An introduction to multivariate statistical
analysis. New York: John Wiley and Sons. 374 pp.
3. Anderson, R. R. and F. J. Wobber. 1973. Wetlands mapping in
New Jersey. Photogrammetric Engineering. 39(4) :353-358.
4. Anonymous. 1973. IPL user documentation for applications pro-
grams. Revision 9. Pasadena: Jet Propulsion Laboratory,
California Institute of Technology. 206 pp.
5. Arvensen, J. C., E. C. Weaver and J. P. Millard. 1971. Rapid
assessment of water pollution by airborne measurement of
chlorophyll. Proceedings Joint Conference on Sensing of
Environmental Pollutants. American Institute of Aeronautics
and Astronautics. Palo Alto. AIAA-71-1097. 7 pp.
6. Atkins, W. R. G. and H. H. Poole. 1952. An experimental study
of the scattering of light by natural waters. Proceedings
of the Royal Society (London). Series B. 140:321-338.
7. Atwell, B. H. and G. C. Thomann. 1972. Mississippi Sound remote
sensing study. Fourth Annual Earth Resources Program Review.
January 17-21, 1972. NASA Manned Spacecraft Center, Houston.
8. Barr, B. G. 1973. Detection of water bodies in Saline County,
Kansas. NASA-CR-132191 . NTIS-E73-10684. 10 May 1973. 4
pp.
9. Beeton, A. M. 1965. Eutrophication of the St. Lawrence Great
Lakes. Limnology and Oceanography. 10(2) :240-254.
10. Beeton, A. M. and W. T. Edmondson. 1972. The eutrophication
problem. Journal Fisheries Research Board of Canada.
29(6):673-682.
11. Berg, C. 0. 1963. Middle Atlantic States. In: Limnology in
North American. Edited by D. G. Frey. Madison: University
of Wisconsin Press, pp. 191-237.
128
-------
12. Billingsley, F. C. and A. F. H. Goetz. 1973. Computer techniques
used for some enhancements of ERTS images. In: Symposium on
Significant Results obtained from the Earth Resources Tech-
nology Satellite-1. 5-9 March 1973. NASA (Goddard Space
Flight Center). Edited by S. C. Freden, E. P. Mercanti, and
M. A. Becker. I(Technical Presentations, Section B):1159-
1167.
13. Birge, E. A. and C. Juday. 1929. Transmission of solar radiation
by the waters of inland lakes. Transactions Wisconsin Acad-
emy Science Arts Letters. 24:509-580.
14. . 1930. A second report on solar radiation and
inland lakes. Transactions Wisconsin Academy Science Arts
Letters. 25:285-335.
15. . 1931. A third report on solar radiation and
inland lakes. Transactions Wisconsin Academy Science Arts
Letters. 26:383-425.
16. . 1932. Solar radiation and inland lakes, fourth
report. Transactions Wisconsin Academy Science Arts Letters.
27:523-562.
17- Black, A. P. and R. F. Christman. 1963. Characteristics of
colored surface water. Journal American Water Works Associ-
ation. 55(6):753-770.
18. Blackwell, R. J. and F. C. Billingsley. 1973. JPL/IPL multispec-
tral photography methods for the detection of algae. Report
730-00048-0-8240. Pasadena: Jet Propulsion Laboratory,
California Institute of Technology. 15 pp.
19. Blackwell, R. F. and D. H. Boland. 1974. Computer analysis of
lakes from ERTS imagery. In: Northwest '74 Pacific Regional
Symposium on Computing the Resources Industry. Canadian
Information Processing Society. 23-24 May 1974. Vancouver,
B. C. pp. 147-157.
20. Boland, D. H. P. 1975. An evaluation of .the Earth Resources
Technology Satellite (ERTS-1) multispectral scanner as a
tool for the determination of lacustrine trophic state.
Ph. D. Thesis. Oregon State University: Corvallis. 311 pp.
21. Boland, D. H. P- and R. J. Blackwell. 1975. The LANDSAT-1
multispectral scanner as a tool in the classification of
inland lakes. In: Proceedings NASA Earth Resources Survey
Symposium. 9-12 June 1975. Houston, Texas. I-A (Technical
Session). NASA TM X-58168. JSC-09930. pp. 419-442.
129
-------
22. Bowker, D. E., P. Fleischer, T. A. Gosink, W. J. Hanna, and J.
Ludwick. 1973. Correlation of ERTS multispectral imagery
with suspended matter and chlorophyll in Lower Chesapeake
Bay. In: Symposium on Significant Results obtained from
the Earth Resources Technology Satellite-1. 5-9 March 1973.
NASA (Goddard Space Flight Center). I(Section B):1291-1297.
23. Bressette, W. E. and D. E. Lear, Jr. 1973. The use of near-
infrared reflected sunlight for biodegradable pollution
monitoring. In: Proceedings Second Conference on Environ-
mental Research Center. Las Vegas, Nevada. Session 1:70-89.
24. Brezonik, P. L. 1969. Eutrophication: the process and its
modeling potential. In: Proceedings Workshop Modeling the
Eutrophication Process. Gainesville: University of Florida.
pp. 68-110.
25. Brezonik, P. L. and E. E. Shannon. 1971. Trophic state of lakes
in north central Florida. Publication 13. Florida Water
Resources Research Center. Gainesville: University of
Florida. 102 pp.
26. Brown. W. L., F. C. Polcyn, J. R. McKimmy, and 0. E. Prewett.
1972. Calculating water quality parameters using remotely
sensed scanner data. In: Proceedings of the Eighth Inter-
national Symposium on Remote Sensing of Environment. 2-6
October 1972. Ann Arbor: University of Michigan. 1:563-567.
27. Carlson, P. R. 1973. Principal sources and dispersal patterns of
suspended particulate matter in nearshore surface waters of
the northeast Pacific Ocean and the Hawaiian Islands. Pro-
gress Report 1 April - 31 May 1973. NASA-CR-133155. NTIS-
#73-10788. 12 June 1973. 9 pp.
28. Chase, P. E. and V. E. Smith. 1973. Utilization of ERTS-1 data
to monitor and classify eutrophication of inland lakes.
Interim Report September 1972 - January 1973. NASA-CR-130555.
NTIS-E73-10299. 70 pp.
29. Chase, P., E. L. Reed, and V. E. Smith. 1973. Utilization of
ERTS-1 data to monitor and classify eutrophication of inland
lakes. In: Symposium of Significant Results obtained from
the Earth Resources Technology Satellite-l. 5-9 March 1973.
NASA (Goddard Space Flight Center). I(Section B):1597-1604.
130
-------
30. Clapp, J. L. 1972. Application of remote sensing to water
resources problems. In: Fourth Annual Earth Resources
Program Review. 17-21 January 1972. NASA Manned Spacecraft
Center, Houston. 2(46):l-39.
31. Clarke, G. L. 1969. The significance of spectral changes in
light scattered by the sea. In: Remote Sensing in Ecology.
Edited by P. L. Johnson. Athens: University of Georgia
Press, pp. 164-172.
32. Clarke, G. L. and H. R. James. 1939. Laboratory analysis of the
selective absorption of light by sea water. Journal Optical
Society America. 29:43-55.
33. Clarke, G. L., G. C. Ewing, and C. J. Lorenzen. 1969. Remote
measurement of ocean color as an index of biological produc-
tivity. A preliminary summary report. In: Proceedings of
the Sixth International Symposium on Remote Sensing of
Environment. 13-16 October 1969. Ann Arbor: University
of Michigan, pp. 991-1001.
34. . 1970. Spectra of backscattered light from the
sea obtained from aircraft as a measure of chlorophyll con-
centration. Science. 167:1119-1121.
35. Coker, A. E., A. Higer, and C. R. Goodwin. 1973. Detection of
turbidity dynamics in Tampa Bay, Florida using multispectral
imagery from ERTS-1. In: Symposium on Significant Results
obtained from the Earth Resources Technology Satellite-l.
5-9 March 1973. NASA (Goddard Space Flight Center). I(Sec-
tion B):1715-1726.
36. Collins, J. R. 1925. Change in the infrared absorption spectrum
of water with temperature. Physics Review. 26:771.
37. Colwell, R. N. 1973. Remote sensing as an aid to the management
of earth resources. American Scientist. 61(2):175-183.
38. Committee on Colorimetry. 1966. The science of color. Sixth
Edition. Washington, D. C.: Optical Society of America.
385 pp.
39. Conrod, A. C. and K. A Rottweiler. 1971. Water quality measure-
ments with airborne multispectral scanners. Paper 71-1096
presented at Joint Conference on Sensing of Environmental
Pollutants. American Institute of Aeronautics and Astro-
nauticSo 8-10 November 1971. Palo Alto, California. 8 pp.
131
-------
40. Crew, G. W. 1973. Remote detection of water pollution with MOCS;
an imaging multispectral scanner. In: Proceedings Second
Conference on Environmental Quality Sensors. 10-11 October
1973. National Environmental Research Center. Las Vegas,
Nevada. Session 11:17-39.
41. Curran, R. J. 1972. Ocean color determination through a scatter-
ing atmosphere. Applied Optics. 11(8):1857-1866.
42. Davis, F. J. 1941. Surface loss of solar and sky radiation by
inland lakes. Transactions Wisconsin Academy Science Arts
Letters. 33:83-93.
43. Davis, W. M. 1882. On the classification of lake basins. Pro-
ceedings Boston Society Natural History. 21:315-381.
44. ___. 1887. The classification of lakes. Science.
10(241) :142-143.
45. Dawson, L. H. and E. 0. Hulburt. 1937. The scattering of light
by water. JournaT Optical Society America. 27:199-201.
46. Deevey, E. S. 1942. Some geographic aspects of limnology. The
Science Monthly. 55:423-434.
47. Detwyler, T. R. 1971. Modern man and environment. In: Man's
Impact on Environment. Edited by T. R. Detwyler. New York:
McGraw-Hill. 731 pp.
48. Donaldson, J. R. 1969. The classification of lakes. In: Pro-
ceedings of the Eutrophication-Biostimulation Assessment
Workshop. 19-21 June 1969. Held at Hotel Claremont,
Berkeley, California. Sponsored by University of California
(Berkeley) and Pacific Northwest Water Laboratory, U. S.
Environmental Protection Agency, pp. 171-185.
49. Draper, N. R. and H. Smith. 1966. Applied regression analysis.
New York: John Wiley and Sons. 407 pp.
50. Duntley, S. Q. 1963. Light in the sea. Journal Optical Society
America. 53:214-233.
51. Dybdahl, A. W. 1973. Detection of dissolved oxygen in water
through remote sensing techniques. Progress Report. In:
Proceedings Second Conference on Environmental Quality
Sensors. 10-11 October 1973. National Environmental
Research Center. Las Vegas, Nevada. Section VIII:48-66.
132
-------
52. Edmondson, W. T. 1974. Review of: Environmental phosphorus
handbook, Edited by E. J. Griffith, A. Beeton, J. M. Spencer,
and D. T. Mitchell. New York: Wiley-Interscience. 1973.
718 pp. In: Limnology and Oceanography. 19(2):369-375.
53. Erb, R. B. 1973. Utilization of ERTS-1 data in the Houston area.
NASA-TM-X-69261. NTIS-E73-10672. 12 March 1973. 2pp.
54. Ewan, T. 1894. On the absorption spectra of dilute solutions.
Proceedings Royal Society (London). 57:117-161.
55. Fore!, F. A. 1889. Ricerche fisiche sui laghi d'Insubria.
Reale Istituto Lombardo di Scienze e Lettere. Rendiconti,
Milano. Second Series. 22:739-742.
56. Fortunatov, M. V. 1957. Aerial methods and their application in
limnological study in inland water bodies. In: Transactions
of the Sixth Conference on the Biology of Inland Waters.
Edited by E. N. Pavlovskii. Original in Russian. Israel
Program for Scientific Translation. Jerusalem: LPST Press.
1969. pp. 591-601.
57. Fraser, A. R. and M. Kovats. 1966. Stereoscopic models of multi-
variate statistical data. Biometrics. 22(2):358-367.
58. Freden, S. C. 1973. Introduction: performance of sensors and
systems. In: Earth Resources Technology Satellite-1
Symposium Proceedings. 29 September 1972. NASA (Goddard
Space Flight Center). Compiled by W. A. Finch, Jr. pp. 1-6.
59. Frieden, H. 1971. Image processing system VICAR guide to system
use. Original 1 October 1968. Revised July 1971. Jet
Propulsion Laboratory. California Institute of Technology.
Pasadena. RO 71-135. 44 pp.
60. Fruh, E. G., K. M. Stewart, G. F. Lee, and G. A. Rohlich. 1966.
Measurements of eutrophication and trends. Water Pollution
Control Federation Journal. 38(8):T237-1258.
61. Gerd, W. V. 1957. An attempt at the biolimnological regionaliza-
tion of the USSR. In: Transactions of the Sixth Conference
on the Biology of Inland Waters. Edited by E. N. Pavlovskii.
Original in Russian. Israel Program for Scientific Trans-
lations. Jerusalem: IPST Press. 1969. pp. 138-145.
62. Gower, J. C. 1966. Some distance properties of latent root and
vector methods used in multivariate analysis. Biometrika.
53:325-338.
133
-------
63. Gower, J. C. 1967. A comparison of some methods of cluster
analysis. Biometrics. 23(4):623-637.
64. Gramms, L. C. and W. C. Boyle. 1971. A literature review of
remote sensing for water quality. Remote Sensing Program
Report Number 7- Madison: University of Wisconsin. 34 pp.
65. Guthrie, D., C. Avery, and K Avery. 1973. Statistical interactive
programming system (*SIPS) user's reference manual. Technical
Report Number 36. Department of Statistics. Corvallis:
Oregon State University. Ill pp.
66. Hammond, E. H. 1964. Analysis of properties in land form geogra-
phy: An application to broad-scale landform mapping. Annals
American Association Geographers. 54:11-23.
67. Hansen, K. 1962. The dystrophic lake type. Hydrobiologia.
19(2):183-191.
68. Harding, S. T. 1942. Lakes, pp. 220-243. In: Hydrology.
Edited by 0. E. Meinzer. New York: McGraw-Hill. Dover
edition by Dover Publications: New York. 712 pp.
69. Hasler, A. D. 1947. Eutrophication of lakes by domestic drainage.
Ecology. 28:383-395.
70. Hasler, A. D. and B. Ingersoll. 1968. Dwindling lakes. Natural
History. 77(9):8-19.
71. Hazen, A. 1892. A new color standard for natural waters. Ameri-
can Chemical Society Journal. 14:300-310.
72. Hidalgo, J. U., A. E. Smalley, K. H. Faller, and M. B. In/in.
1973. Preliminary study of Lake Pontchartrain and vicinity
using remotely sensed data from the ERTS-A satellite.
Progress Report. NASA-CR-132010. NTIS-E73-10613. 15 May
1973. 6 pp.
73. Hoffer, R. M., P- E. Anuta, and T. L. Phillips. 1972. ADP,
multiband, and multiemulsion digitized photos. Photogram-
metric Engineering. 38(10):989-1001.
74. Horn, L. W. 1968. Remote sensing of water pollution. Journal
Water Pollution Control Federation. 40:1728.
75. Holmes, R. W. 1970. The Secchi disk in turbid coastal waters.
Limnology and Oceanography. 15(5):688-694.
134
-------
76. Hooper, F. F. 1969. Eutrophication indices and their relation to
other indices of ecosystem change. In: Eutrophication:
Causes, Consequences, Correctives. Proceedings of a Sym-
posium. 11-15 June 1967. University of Wisconsin. Washing-
ton, D. C.: National Academy of Science, pp. 225-235.
77. Horan, J. J., D. S. Schwartz, and J. D. Love. 1974. Partial
performance degradation of a remote sensor in a space
environment, and some probable causes. Applied Optics.
13(5):1230-1237.
78. Hotelling, H. 1933a. Analysis of a complex of statistical
variables into principal components (I, Introduction). The
Journal of Educational Psychology. 24:417-441.
79. . 1933b. Analysis of a complex of statistical
variables into principal components (II). The Journal of
Educational Psychology. 24:498-520.
80. . 1936. Simplified calculation of principal com-
ponents. Psychometrika. l(l):27-35.
81. Hulburt, E. 0. 1945. Optics of distilled and natural water.
Journal of the Optical Society of America. 35(11):698-705.
82. Hunt, C. B. 1967. Physiography of the United States. San
Francisco: W. H. Freeman. 480 pp.
83. Hulstrom, R. L. 1973. The cloud bright spot. Photogrammetric
Engineering. 39(4):370-376.
84. Hutchinson, 6. E. 1957. A treatise on limnology. Volume I.
Geography, physics, and chemistry. New York: John Wiley.
1015 pp.
85. . 1967. A treatise on limnology. Volume II.
Introduction to lake biology and the limnoplankton. New
York: John Wiley. 1115 pp.
86. . 1973. Eutrophication. The scientific background
of a contemporary practical problem. American Scientist.
61:269-279.
87. James, H. R. and E. A. Birge. 1938. A laboratory study of the
absorption of light by lake waters. Transactions Wisconsin
Academy Science Arts Letters. 31:1-154.
135
-------
88. James, H. R. 1941. Beer's law and the properties of organic
matter in lake waters. Transactions Wisconsin Academy
Science Arts Letters. 33:73-82.
89. Jarnefelt, H. 1958. On the typology of northern lakes. In:
Proceedings International Association Theoretical Applied
Limnology (International Vereinigung fur Theoretische und
Angewandte Limnologie). Volume XIII. Congress in Finland
1956. Edited by T. T. Macan. Stuttgart: 1958. pp. 228-235.
90. Jerlov, N. G. 1968. Optical oceanography. Elsevier Oceanography
Series. Volume 5. Amsterdam: Elsevier. 194 pp.
91. Johnson, S. C. 1967. Hierarchical clustering schemes. Psycho-
metrika. 32(3):241-254.
92. Juday, C. and E. A. Birge. 1933. The transparency, the color,
and the specific conductance of the lake water of northeastern
Wisconsin. Transaction Wisconsin Academy Science Arts
Letters. 28:205-259.
93. Katzoff, S. 1962. The electromagnetic-radiation environment of a
satellite, Part 1, range of thermal to X-radiation. NASA TN
D-1360. N62-15249. 75 pp.
94. Kelly, M. G. and A. Conrod. 1969. Aerial photographic studies of
shallow water benthic ecology. In: Remote Sensing in
Ecology. Athens: University of Georgia Press, pp. 173-184.
95. Ketelle, M. J. and P. D. Uttormark. 1971. Problem lakes in the
United States. Water Pollution Control Research Series.
Report Number 16010 EHR. U. S. Environmental Protection
Agency. 282 pp.
96. Kiefer, R. W. and J. P. Scherz. 1970. Applications of airborne
remote sensing technology. Journal Surveying Mapping Divi-
sion. Proceeding American Society Civil Engineers.
SU1(7225):57-80.
97. . 1971. Aerial photography for water resources
studies. Journal Surveying Mapping Division. Proceedings
American Society Civil Engineers. SU2(8545):321-333.
98. Klemas, V. 1973. Applicability of ERTS-1 imagery to the study of
suspended sediment and aquatic fronts. In: Symposium on
Significant Results obtained from the Earth Resources
Technology Satellite-1. NASA (Goddard Space Flight Center).
I(Section B):1725-1290.
136
-------
99. Kolipinski, M. C. and A. L. Higer. 1968. Applications of aerial
photography and remote sensing to hydrobiological research in
south Florida. Open File Report. U. S. Geological Survey.
Water Resources Division. Miami, Florida. 7 pp.
100. Kritikos, H., L. Yorinks, and H. Smith. 1973. Water Quality
analysis using ERTS-A data. In: Proceedings Second Confer-
ence on Environmental Quality Sensors. 10-11 October 1973.
National Environmental Research Center. Las Vegas, Nevada.
p. VII:37-56.
101. Lance, G. N. and W. T. Williams. 1967a. A general theory of
classificatory sorting strategies I. Hierarchic systems.
The Computer Journal. 9:373-380.
102. . 1967b. A general theory of classificatory sorting
strategies II. Clustering systems. The Computer Journal.
10:271-277.
103. Larkin, P. A. and T. G. Northcote. 1958. Factors in lake typo-
logy in British Columbia, Canada. In: Proceedings
International Association Theoretical Applied Limnology (Verh.
Internat. Ver. Limnol.). 13:252-263.
104. Lind, A. 0. 1973. Survey of lake flooding from ERTS-1: Lake
Champlain. NASA-CR-133138. NTIS-#72-10771. June 1973.
H PP-
105. Lind, A. 0. and E. B. Henson. 1973. Pollution monitoring in
Lake Champlain using ERTS-1 imagery. NASA-CR-133086.
NTIS-E73-10742. June 1973. 8 pp.
106. Lindeman, R. L. 1942. The trophic-dynamic aspect of ecology.
Ecology. 23(4):399-418.
107. Look, D. C., A. L. Crosbie, B. F. Armaly, and H. F. Nelson. 1973.
Influence of absorption coefficient on reflectance of water.
Journal Water Pollution Control Federation. 45:936-939.
108. Lorenzen, C. J. 1970. Survey chlorophyll as an index of the
depth, chlorophyll content, and primary productivity of the
euphotic layer. Limnology and Oceanography. 15(3):479-480.
109. Lueschow, L. A., J. M. Helm, D. R. Winter, and G. W. Karl. 1970.
Trophic nature of selected Wisconsin lakes. Wisconsin
Academy Sciences Arts Letters. 58:237-264.
137
-------
110. McKeon, J. J. 1967. Hierarchical cluster analysis computer pro-
gram. Biometrics Laboratory. George Washington University,
Kensington, Maryland. 6 pp.
111. Margalef, R. 1958. "Trophic" typology versus biotic typology as
exemplified in the regional limnology of northern Spain. In:
Proceedings International Association Theoretical Applied
Limnology. 13:339-349.
112. Maugh, T. H. 1973. ERTS: Surveying earth's resources from
space. Science. 180(4081):49-1044.
113. . 1973. ERTS (II): A new way of viewing the
earth. Science. 180(4082):171-173.
114. Morrison, D. F. 1967. Multivariate statistical methods. New
York: McGraw-Hill. 338 pp.
115. Moss, W. W. 1967. Some new analytic and graphic approaches to
numerical taxonomy, with an example for the Dermanyssidae
(Acari). Systematic Zoology. 16:177-207.
116. . 1968. Experiments with various techniques of
numerical taxonomy. Systematic Zoology. 20:309-330.
117. Moyle, J. B. 1945. Classification of lake waters upon the basis
of hardness. Proceedings Minnesota Academy Science. 13:8-12.
118. . 1946. Some indices of lake productivity. Trans-
actions of the American Fisheries Society. 76:322-334.
119. Minnesota Department of Conservation. 1968. Inventory of
Minnesota lakes. Bulletin Number 25. Division Water, Soils,
and Minerals. 498 pp.
120. Mueller, J. L. 1972. Remote measurement of chlorophyll concen-
tration and Secchi-depth using the principal components of
the ocean's color spectrum. In: Fourth Annual Earth
Resources Program Review. 17-21 January 1972. NASA
Manned Spacecraft Center, Houston. 4(105):1-13.
121. Nace, R. L. 1960. Water management, agriculture, and ground-
water supplies. U. S. Geological Survey Circular 415. 11 pp.
122. National Aeronautics and Space Administration. 1972a. NASA earth
resources technology satellite data users handbook. Goddard
Space Flight Center. Document Number 71SD4249. Subject to
revision.
138
-------
123. National Aeronautics and Space Administration. 1972b. ERTS
investigators' bulletin. A(4). 8 August 1972. 1 pp.
124. Naumann, E. 1919. Nagra synpunkte angaende planktons okologi.
Med. sarskild ha'nsyn till fytoplanton. Svensk Botanisk
Tidskrift. 13:129-158.
125. . 1931. Limnologische terminologie. Urban and
Schwarzenberg, Berline-Wien. (pp. 153 and 413). 776 pp.
126. National Eutrophication Survey. 1974. National Eutrophication
Survey methods for lakes sampled in 1972. U. S. Environ-
mental Protection Agency National Eutrophication Survey
Working Paper Number 1. PNERL (NERC-Corvallis) and NERC-Las
Vegas. 40 pp.
127. Nielsen) U. 1972. Agfacontour film for interpretation. Photo-
grammetric Engineering. 38(1):1009-1104.
128. Nordberg, W. 1972. Written communication dated 9 November 1972.
NASA Goddard Space Flight Center. Greenbelt, Maryland. 3 pp.
129. Padron, M. 1969. An axiomatic basis and computational methods
for optimal clustering. Technical Report 18. Department of
Industrial and Systems Engineering. University of Florida:
Gainesville. 171 pp.
130. Pearsall, W. H. 1932. Phytoplankton in the English lakes. II.
The composition of the phytoplankton in relation to dissolved
substances. Journal of Ecology. 20(2):241-262.
131. Pennak, R. W. 1958. Regional lake typology in northern Colorado,
U. S. A. In: Proceedings International Association Theoreti-
cal Applied Limnology. 13:262-283.
132. Piech, K. R. and J. E. Walker. 1971. Aerial color analyses of
water quality. Journal Survey Mapping Division. Proceedings
American Society Civil Engineers. SU2(8507):185-197.
133. Pietenpol, W. B. 1918. Selective absorption in the visible
spectrum of Wisconsin waters. Transactions Wisconsin Academy
Science Arts Letters. 19:562.
134. Pluhowski, E. J. 1973. Remote sensing of turbidity plumes in Lake
Ontario. In: Symposium on Significant Results obtained from
the Earth Resources Technology Satellite-!. NASA (Goddard
Space Flight Center). 5-9 March 1973. I(Section B):837-846.
139
-------
135. Polcyn, F. C. and D. R. Lynzenga. 1973. Calculations of water
depth from ERTS-MSS data. In: Symposium on Significant
Results obtained from the Earth Resources Technology
Satellite-!. 5-9 March 1973. NASA (Goddard Space Flight
Center). I(Section B):1433-1441.
136. Raman, C. V. 1922. On the molecular scattering of light in
water and the colour of the sea. Proceedings Royal
Society (London). Ser. A. 101:64-79.
137. Ranz, E. and S. Schneider. 1970. Der Aquidensitenfilm als
Hilfsmittel bei der Photointerpretation. Bildmessung und
Luftbildwesen. 2:123-134.
138. Rawson, D. S. 1956. Algal indicators of trophic lake types.
Limnology and Oceanography. l(l):18-25.
139. . 1960. A limnological comparison of twelve
large lakes in northern Saskatchewan. Limnology and
Oceanography. 5(2):195-211.
140. Reeves, C. C. 1973. Dynamics of playa lakes in the Texas high
plains. Symposium on Significant Results Obtained from the
Earth Resources Technology Satellite-l. 5-9 March 1973.
NASA (Goddard Space Flight Center). I (Section B)-.809-817.
141. Robinove, C. J. 1965. Infrared photography and imagery in water
resources research. Journal American Water Works Association.
57(7):834-840.
142. Rodhe, W. 1969. Crystallization of eutrophication concepts in
northern Europe. In: Eutrophication: Causes, Consequences,
Correctives. Proceedings of a Symposium. 11-15 June 1967.
University of Wisconsin. Washington, D. C.: National
Academy of Science, pp. 50-64.
143. Rogers, R. H. and K Peacock. 1973. A technique for correcting
ERTS data for solar and atmospheric effects. In: Sym-
posium on Significant Results obtained from the Earth
Resources Technology Satellite-l. 5-9 March 1973. NASA
(Goddard Space Flight Center). I(Section B):1115-1122.
144. Roger, R. H. and V. E. Smith. 1973. Utilization of ERTS-1 data
to monitor and classify eutrophication of inland lakes.
Interim Report for February 1973-July 1973. NTIS-E73-10898.
117 pp.
145. Rohlf, F. J. 1968. Stereograms in numerical taxonomy. Systematic
Zoology. 17:246-255.
140
-------
146. Rohlf, F. J. 1968. Grafpac, Graphic output subroutines for the
GE 635 computer. Computer Contribution Number 36. Edited
by D. F. Merriam. State Geological Survey. University of
Kansas: Lawrence. 50 pp.
147. Ross, D. S. 1973a. Water depth estimation with ERTS-1 imagery.
In: Symposium on Significant Results obtained from the
Earth Resources Technology Satellite-1. 5-9 March 1973.
NASA (Goddard Space Flight Center). I(Section B):1423-1432.
148. . 1973b. Atmospheric effects in multispectral
photographs. Photogrammetric Engineering. 39(4):377-384.
149. Rossolimo, L. L. 1971. Anthropogenic eutrophication of lakes and
reservoirs: nature of the problem and research goals.
Hydrobiological Journal. 7(3):90-97.
150. Round, F. E. 1958. Algal aspects of lake typology. In: Pro-
ceedings International Association Theoretical Applied
Limnology. 13:306-310.
151. Russell, I. C. 1895. Lake of North America. A reading lesson for
students of geography and geology. Boston: Gin and Company.
125 pp.
152. Ruttner, F. 1963. Fundamentals of limnology. Third edition.
Translated by D. G. Frey and F. E. J. Fry. Toronto:
University of Toronto Press. 295 pp.
153. Sawyer, C. N. 1946. Fertilization of lakes by agricultural and
urban drainage. Journal New England Water Works Association.
61:109-127.
154. Sawyer, W. R. 1931. The spectral absorption of light by pure
water and Bay of Fundy water. Contribution to Canadian
Biology and Fisheries being Studied from the Biological
Stations of Canada. Series D. Hydrographic. 7(8):73-89.
155. Scherz, J. P., D. R. Graff, and W. C. Boyle. 1969. Photographic
characteristics of water pollution. Photogrammetric
Engineering. 35:38-43.
156. Scherz, J. P. 1971a. Monitoring water pollution by means of remote
sensing techniques. Remote Sensing Program Report Number 3.
Madison: University of Wisconsin. 27 pp.
157. . 1971b. Monitoring water pollution by remote
sensing. Journal Surveying Mapping Division. Proceedings
American Society Civil Engineers. 97:307-320.
141
-------
158. Schmer, F. A., M. J. Tipton, D. W. Ryland, 0. Hayden, and G.
Beaver. 1972. Investigation of lake water quality in
eastern South Dakota. In: Proceedings Eighth International
Symposium Remote Sensing Environment. 2-6 October 1972. Ann
Arbor: University of Michigan. 1:553-562.
159. Schnell, G. D. 1970a. A phenetic study of the suborder Lari
(Aves)/I. Methods and results of principal components
analyses. Systematic Zoology. 19:35-37.
160. . 1970b. A phenetic study of the suborder Lari
(Aves)/II. Phenograms, discussion, and conclusions.
Systematic Zoology. 19:264-302.
161. Shannon, E. E. 1970. Eutrophication-trophic state relationships
in north and central Florida lakes. Doctoral dissertation.
Gainesville: University of Florida. 257 pp.
162. Shannon, E. E. and P. L. Brezonik. 1972a. Eutrophication
analysis: a multivariate approach. Journal Sanitary
Engineering Division. Proceedings American Society Civil
Engineers. 98(SA1, 8735):35-37.
163. . 1972b. Relationships between trophic state and
nitrogen and phosphorus loading rates. Environmental
Science and Technology. 6(8):719-725.
164. Seal, H. L. 1964. Multivariate statistical analysis for biol-
ogists. London: Methuen. 207 pp.
165. Seddon, B. 1972. Aquatic macrophytes as limnological indicators.
Freshwater Biology. 2:107-130.
166. Sheldon, A. 1972. A quantitative approach to the classification
of inland waters. In: Natural Environments. Studies in
Theoretical and Applied Analysis. Edited by J. V. Krutilla.
Baltimore: John Hopkins University Press, pp. 205-261.
167. Silvestro, F. B. 1970. Remote sensing analysis of water quality.
Journal Water Pollution Control Federation. 40:553-561.
168. Smith, R. C., J. E. Tyler, and C. R. Goldman. 1973. Optical
properties and color of Lake Tahoe and Crater Lake. Limnology
and Oceanography. 18(2):189-199.
169. Sneath, P. H. A. and R. R. Sokal. 1973. Numerical taxonomy:
The principles and practice of numerical classification. San
Francisco: F. H. Freeman. 573 pp.
142
-------
170. Snedecor, G. W. and W. G. Cochran. 1967. Statistical methods.
Sixth edition. Ames: Iowa State University Press. 593 pp.
171. Soil Survey Staff. 1960. Soil Classification: A comprehensive
system, 7th approximation. U. S. Department of Agriculture.
Soil Conservation Service. 265 pp.
172. . 1967. Supplement to soil classification system
(7th approximation). U. S. Department of Agriculture. Soil
Conservation Service. 207 pp.
173. Sokal, R. R. and F. J. Rohlf. 1962. The comparison of dendo-
grams by objective methods. Taxon. 11(2):33-40.
174. Sokal, R. R. and P. H. A. Sneath. 1963. Principals of numerical
taxonomy. San Francisco: W. H. Freeman. 359 pp.
175. Specht, M. R., D. Needier, and N. L. Fritz. 1973. New color
film for water-photography penetration. Photogrammetric
Engineering. 39(4):359-369.
176. Spiess, F. N. 1970. Oceanic environment. In: Hydronautics.
Edited by: H. E. Sheets and V. T. Boatwright, Jr. New
York: Academic Press.
177. Stewart, K. M. and G. A. Rohlich. 1967. Eutrophication -- A
review. Publication Number 35. California State Water
Quality Control Board. Sacramento. 188 pp.
178. Strahler, A. N. 1969. Physical geography. Third edition. New
York: John Wiley. 733 pp.
179. Strong, A. E. 1973. ERTS-1 observes algal blooms in Lake Erie
and Utah Lake. Symposium on Significant Results obtained
from the Earth Resources Technology Satellite-1. 5-9 March
1973. NASA (Goddard Space Flight Center). I(Section B):
1605-1612.
180. Szekielda, K. H. and R. J. Curran. 1972. Chlorophyll structure
in the ocean. NASA-72-31338. NTIS-E72-10047. 29 September
1972. 2 pp.
181. Thienemann, A. 1918. Untersuchungen uber die Beziechungen
zwischen dem Sauerst offgehalt der Wassers und der Zusammen-
setzung der Fauna in norddeutchen Seen. Archiv Fuer Hydro-
biologie. 12:1-65.
182. Thornthwaite, C. W. 1948. An approach toward a rational classifi-
cation of climate. Geographical Review. 38:55-940
143
-------
183. Trewartha, G. T. 1968. An introduction to climate. New York:
McGraw-Hill. 408 pp.
184 Tyler, J. E. 1968. The Secchi disc. Limnology and Oceanography.
13(0:1-6.
185. Ule, W. 1892. Die Bestimnumg der Wasserfarbe in den Seen. A.
Petermanns Mittheilungen aus Justus Perthes' Geographisher
Anstalt. 38:70-71.
186. U. S. Environmental Protection Agency. 1971. Algal assay pro-
cedure bottle test. National Eutrophication Research Pro-
gram. Pacific Northwest Water Laboratory. Corvallis. 82 pp.
187. U. S. Department of Agriculture. 1941. Climate and Man. Year-
book of Agriculture. Washington, D. C.: U. S. Government
Printing Office. 1248 pp.
188 U. S. Geological Survey. 1965. Geological map of North America.
1:5,000,000.
189. o 1970. The national atlas of the United States.
Washington, D. C.: U. S. Government Printing Office. 417 pp.
190. Van Lopik, J. R., G. S. Rombie, and A. E. Pressman. 1968.
Pollution surveillance by noncontact infrared techniques.
Journal Water Pollution Control Federation. 40(3):425-438.
191. Veatch, J. 0. and C. R. Humphrys. 1966. Water and water use
terminology. Kaukauna: Thomas Printing and Publishing.
375 pp.
192. Visher, S. S. 1954. Climatic atlas of the United States.
Cambridge: Harvard University Press. 403 pp.
193. Vollenweider, R. A. 1968. Scientific fundamentals of the eutro-
phication of lakes and flowing waters with particular refer-
ence to nitrogen and phosphorus as factors in eutrophication.
Technical report prepared for the Organization for Economic
Cooperation and Development. Paris, France. 159 pp.
194. Ward, J. H. Jr. 1963. Hierarchical grouping to optimize an
objective function. American Statistical Association Journal.
58:236-244.
195. Watanabe, K. 1973. Polluted and turbid water masses in Osaka Bay
and its vicinity revealed with ERTS-A imageries. Symposium
on Significant Results obtained from the Earth Resources
Technology Satellite-!. 5-9 March 1973. NASA (Goddard Space
Flight Center). I(Section B):681-688.
144
-------
196. Weber, C. A. 1907- Aufbau and Vegetation der Moore Norddeutsch-
lands. Beiblatter Botanische Jahrbucker fur Systematik,
Pflanzengeschichte und Pflanzengeographie. 90:19-34.
Supplement to Bot. Jahrb. 40.
197. Welch, P. S. 1952. Limnology. Second edition. New York:
McGraw-Hill. 538 pp.
198. Wezernak, C. T. and F. C. Polcyn. 1972. Eutrophication assess-
ment using remote sensing techniques. Proceedings of the
Eighth International Symposium on Remote Sensing of Environ-
ment. 2-6 October 1972. Ann Arbor: University of Michigan.
1:541-551.
199. Whipple, G. C. 1898. Classifications of lakes according to
temperature. American Naturalist. 32(373):25-33.
200. White, P0 G. 1969. Experimental results of the remote measure-
ment of ocean color. In: Second Annual Earth Resources
Aircraft Program Status Review. 16-18 September 1969.
NASA Manned Spacecraft Center. Houston. 3(50):1-9.
201. Williams, L. 0. and J. D. Samol. 1968. Effects on photographs of
water by variation of sediment and algal content under
controlled conditions. Technical Report Number 9. Office
of Naval Research Contract Number N00014-67-A0102-0001.
Johnson City: East Tennessee State University Remote Sensing
Institute. 6 pp. + 6 pp. photographs.
202. Williams, W. T. and M. B. Dale. 1965. Fundamental problems in
numerical taxonomy. Advances in Botanical Research. 2:35-68.
203. Winner, R. W. 1972. An evaluation of certain indices of eutrophy
and maturity in lakes. Hydrobiologia. 40(2):223-245.
204. Work, E. A. Jr., D. S. Gilmer, and A. T. Klett. 1973. Preliminary
evaluation of ERTS-1 for determining numbers and distribution
of prairie ponds and lakes. In: Symposium on Significant
Results obtained from the Earth Resources Technology
Satellite-!. 5-9 March 1973. NASA (Goddard Space Flight
Center). I(Section B):801-808.
205. Wright, F. F., G. D. Sharma, and D. C. Burbank. 1973. ERTS-1
observations of sea surface circulation and sediment trans-
port, Cook Inlet, Alaska. In: Symposium on Significant
Results obtained from the Earth Resources Technology
Satellite-l. 5-9 March 1973. NASA (Goddard Space Flight
Center). I(Section B) :1315-1322.
145
-------
206. Yarger, H. L., J. R. McCauley, G. W. James, L. M. Magnuson, and
G. Re Marzolf. 1973. Water turbidity detection using ERTS-1
imagery. In: Symposium on Significant Results obtained from
the Earth Resources Technology Satellite-1. 5-9 March 1973.
NASA (Goddard Space Flight Center). I(Section B):651-658.
207. Yen.tch, C. S. 1960. The influence of phytoplankton pigments on
the color of sea water. Deep Sea Research. 7:1-9.
208. . 1962. Measurement of visible light absorption
by particulate matter in the ocean. Limnology and Oceanog-
raphy. 7:207-217.
209. Yentch, C. S. and D. W. Menzel. 1963. A method for the deter-
mination of phytoplankton chlorophyll and phaeophytin by
fluorescence. Deep Sea Research. 10:221-231.
210. Zafar, A. R. 1959. Taxonomy of lakes. Hydrobiologia. 13(3):
287-299.
211. Zumberge, J0 H. 1952. The lakes of Minnesota: their origin and
classification. Minnesota Geological Survey Bulletin
Number 35. 99 pp.
146
-------
SECTION X
LIST OF ACRONYMS AND SYMBOLS
AAY: algal assay yield
ADP: automatic data processing
AIP: automatic image processing
A-space: attribute space
b&w: black and white
bit: binary digit
bpi: bits per inch
CRT: cathode ray tube
CCT: computer-compatible tape
CDC: Control Data Corporation
CHLA: chlorophyll a_
COND: conductivity
cm: centimeter
°C: degrees Celsius
DN: digital number
DCS: data collection system
EBR: electron beam recorder
EPA: Environmental Protection
Agency (United States)
EROS: Earth Resources Observa-
tion Systems
ERTS-1: Earth Resources Tech-
nology Satellite Number
One
ESB: Eutrophication Survey Branch
GRN: Green (MSS Band 4)
GRNIR1: ratio of green to infrared-
one
GRNIR2: ratio of green to infrared-
two
GRNRED: ratio of green to red
GRAFPAC: standard graphic output
subroutines
GSFC: Goddard Space Flight Center
ha: hectare
hz: hertz
IBM: International Business Machines
IFOV: instantaneous field of view
IPL: Image Processing Laboratory
(Jet Propulsion Laboratory)
IR1: infrared-one
IR2: infrared-two
IR1IR2: ratio of infrared-one to
infrared-two
ISEC: inverse of Secchi disc
transparency
I-space: individual space
JPL: Jet Propulsion Laboratory
km: kilometer
1: liter
147
-------
LANDSAT-1: Land Satellite One
tt.e., ERTS-1)
LN: natural logarithm
LNAAY: natural log-transformed
algal assay yteld
LNCHLA: natural log-transformed
chlorophyll a^
LNCOND: natural log-transformed
conductivity
NIR: near-infrared
nm: nanometers
OD: optical density
OS: operating system
OSI: Optimum Systems Incorporated
OSU: Oregon State University
PCA: principal components anaylsis
PCI: principal component trophic
state index
LNISEC: natural log-transformed
inverse of Secchi disc
transparency
pixel: picture element
LNSECCHI: natural log-transformed
Secchi disc transparency PMT: photomultiplier tube
LNTON: natural log-transformed
total organic nitrogen
LNTPHOS: natural log-trans-
formed total phos-
phorus
m: meter
MCR: mean composite rank
yg: microgram
ysec: microsecond
mg: milligram
MSS: multispectral scanner
N: number of observations
NASA: National Aeronautics and
Space Administration
NERC: National Environmental
Recearch Center
NES: National Eutrophication
Survey
PNERL: Pacific Northwest Environ-
mental Research Laboratory
RBV: return beam vidicon
RED: red (MSS Band 5)
R-space: row space (i.e., attribute
space)
REDIR1: ratio of red to infrared-
one
REDIR2: ratio of red to infrared-
two
SECCHI: Secchi disc transparency
SIPS: Statistical Interactive
Programming System
S-matrix: similarity matrix
STORE!: Storage and Retrieval
number
TON: total organic nitrogen
TPHOS: total phosphorus
148
-------
USGS: United States Geological
Survey
VFC: video film converter
VICAR: Video Communication and
Retrieval
XGRNIR1: mean of green to infrared-
one ratios
XGRNIR2: mean of green to infrared-
two ratios
XGRNRED: mean of green to red
ratios
XIR1IR2: mean of infrared-one to
infrared-two ratios
XREDIR1: mean of red to infrared-
one ratios
XREDIR2: mean of red to infrared-
two ratios
A: Euclidian distance
A2: squared Euclidian distance
/\: predicted value
149
-------
SECTION XI
APPENDICES
A. Trophic Indicator Data for 100 NES-sampled Lakes
B. Sampling Dates for 100 NES-sampled Lakes (1972)
C. Morphometry and Hydrology of Study Lakes
D. LANDSAT-1 MSS Models, Concatenations, Area! Relationships, and
Descriptive Statistics
E. N x N Squared Euclidian Distance Matrix
F. Listing of the McKeon Cluster Analysis Program
150
-------
APPENDIX A
TROPHIC INDICATOR DATA FOR 100 NES-SAMPLED LAKES
Lake Serial
Name Number
Blackduck
Bemidji
Andrusia
Wolf
Cass
Leech
Birch
Trout
Mashkenode
Whitewater
Pelican
Shagawa
Gull
Rabbit
Cranberry
Darling
Carlos
Le Homme
Dieu
Minnewaska
Nest
Green
Wagonga
Clearwater
Mud (at
Maple Lake)
Cokato
Buffalo
Carrigan
Silver
Minnetonka
Forest
White Bear
St. Croix
Spring
Pepin
Madison
Sakatah
Bear
Albert Lea
Yellow
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
SECCHI
1.
2.
1.
1.
1.
2.
2.
2.
1.
1.
1.
1.
2.
2.
1.
2.
2.
2.
1.
1.
2.
0.
1.
0.
1.
1.
0.
0.
1.
2.
3.
1.
0.
0.
0.
1.
0.
0.
1.
72
15
37
19
88
17
44
81
05
71
71
77
32
71
18
56
74
31
71
36
79
29
57
35
34
27
28
25
38
16
45
11
47
96
86
89
28
19
49
ISEC
0,
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
3.
0.
2.
0.
0.
3.
3.
0.
0.
0.
0.
2.
1.
1.
0.
3.
5.
0.
58
47
73
84
53
46
41
36
95
59
59
57
43
37
85
39
37
43
59
74
36
41
64
88
54
79
58
94
73
46
29
90
15
05
16
53
58
37
67
COND
244
315
276
384
272
262
205
325
301
139
88
71
204
276
107
391
353
315
638
353
353
808
362
530
540
597
595
470
361
273
253
159
450
417
305
389
431
650
174
TPHOS
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1.
0.
0.
1.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1.
0.
051
049
035
068
027
020
024
045
100
066
033
084
031
036
038
019
014
022
035
050
051
940
033
893
208
224
164
540
070
024
019
054
237
187
064
324
191
002
071
TON
0. 80
0.69
0. 58
0.48
0. 53
0. 53
0. 42
0. 43
0. 56
0. 67
0. 39
0. 53
0. 51
1. 25
0. 78
0. 80
0. 56
0. 67
0.90
1. 03
0,46
1. 54
0. 77
1. 31
1. 39
1. 65
1. 81
3. 35
1. 45
0. 61
1. 32
0. 57
0. 64
2. 13
1. 24
1. 21
1. 57
2. 50
0. 43
CHLA
14.6
9. 5
13. 0
17. 2
9.8
6.2
7. 1
7. 0
25. 3
9.8
11. 4
11. 3
12. 5
6.7
30. 2
11. 8
4.6
12.4
7.6
21.4
4.9
94. 5
12. 7
132. 3
10. 7
38. 1
84. 3
126. 1
16.6
10. 5
5. 2
10. 2
21. 8
14.9
30. 4
10. 8
61. 2
381. 2
13. 7
AAY
2. 2
1. 3
1.8
1.6
0. 8
0. 5
1. 4
1.6
2.6
2.9
1. 3
4. 5
2. 4
2. 0
0. 2
0. 8
0.9
0. 2
0. 4
3. 8
1. 3
13.9
2. 4
8. 5
29.0
16. 2
13. 2
25.9
1. 2
0. 2
0.3
3.4
34. 5
14. 0
3. 8
14. 8
5. 5
61. 3
3. 3
151
-------
APPENDIX A (continued)
TROPHIC INDICATOR DATA FOR 100 NES-SAMPLED LAKES
Lake Serial
Name Number SECCHI ISEC COND TPHOS TON CHLA AAY
Wapogasset
Long
Elk
Trout
Crystal
Tainter
Shawano
Poygan
Butte des
Morts
Winnebago
Round
Green
Swan
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac La
Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Charlevoix
Higgins
Houghton
Pere
Marquette
White
Muskegon
Fremont
Mona
Crystal
Jordan
Thornapple
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
1.
0.
0.
4.
8.
1.
2.
0.
0.
0.
3.
5.
1.
0.
1.
2.
1.
2.
3.
2.
1.
1.
1.
0.
2.
3.
1.
0.
3.
3.
6.
2.
1.
2.
1.
1.
1.
3.
1.
1.
80
88
93
12
03
39
00
48
58
55
40
78
91
43
10
32
58
04
34
49
84
30
72
58
78
59
26
52
22
78
21
01
31
09
61
48
23
40
84
45
0.
1.
1.
0.
0.
0.
0.
2.
1.
1.
0.
0.
0.
2.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1.
0.
0.
0.
1.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
56
13
07
24
13
72
50
07
73
82
29
17
53
34
91
43
63
49
30
40
54
77
58
71
36
28
79
94
31
27
16
50
77
48
52
68
82
29
54
69
198
72
66
96
50
173
220
306
289
316
323
386
378
422
403
378
580
434
437
446
308
503
435
599
435
399
451
390
386
289
231
236
438
442
347
486
446
327
416
576
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
053
046
140
042
007
115
021
077
063
156
018
043
158
388
140
019
333
013
015
016
026
135
036
479
021
012
144
064
015
007
040
015
035
031
081
482
428
009
187
050
0.45
0,87
0.83
0. 24
0. 15
0. 50
0.66
1. 09
0. 70
1. 02
0. 60
0.43
0. 71
0. 31
0. 73
0.63
1.29
0. 57
0.93
0. 70
0. 67
0. 77
1.63
0.97
0. 68
0. 39
1. 02
1. 05
0. 59
0. 19
0. 13
0. 52
0. 29
0.42
0.61
3. 35
0. 79
0. 62
0. 84
0. 71
16.6
7. 1
7. 1
2. 7
1. 5
13. 7
11.9
19.4
25.4
48.4
3. 5
4.8
8. 2
69. 5
30.9
8. 1
36. 1
7.9
3. 1
8.4
7. 5
12.0
15. 5
44. 7
6.4
4. 7
43.9
36. 4
5. 8
3. 0
1. 1
9. 2
11. 8
9. 2
9. 5
28. 5
27. 3
2 9
*-> • /
20. 5
14.7
5. 5
0. 2
2.8
0.7
0.2
16. 8
0. 7
8.9
6.7
13.4
0.2
1. 0
17. 2
12. 7
8.0
0. 3
29. 0
0. 2
0. 2
3. 5
1. 4
21. 8
2. 1
26. 0
0. 8
0. 2
19. 9
0. 3
0 3
^ • *J
0. 1
0. 1
0. 2
5. 5
4 5
~ • — t
7 3
I • *J
44 0
^t~» v
31 6
~J -L • \J
0 1
\J • 1
? ? ?
t-iLit Li
4. 5
152
-------
APPENDIX A (continued)
TROPHIC INDICATOR DATA FOR 100 NES-SAMPLED LAKES
Lake Serial
Name Number
Strawberry
Chemung
Thompson
Ford
Union
Long
Randall
Schroon
Black
Cassadaga
Chautauqua
Conesus
Canandaigua
Keuka
Seneca
Cayuga
Owasco
Cross
Otter
Round
Saratoga
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
SECCHI
1.
2.
2.
1.
1.
1.
1.
3.
1.
2.
2.
3.
4.
3.
4.
2.
2.
1.
1.
1.
2.
97
43
34
11
13
93
08
73
86
57
00
19
33
57
14
80
71
36
14
17
52
ISEC
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
51
41
43
90
88
52
92
27
54
39
50
31
23
28
24
36
37
74
88
86
40
COND
405
402
490
539
510
452
518
51
116
204
152
341
318
246
778
442
280
641
283
262
227
TPHOS
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
168
037
040
111
055
197
203
005
036
029
036
022
014
010
012
015
010
127
078
073
028
TON
0. 75
0. 83
0. 83
0.81
0. 51
1. 00
0. 78
0. 16
0.67
0. 36
0. 41
0. 36
0. 08
0. 18
0. 11
0. 23
0. 19
0. 50
0. 63
0. 58
0. 51
CHLA
11. 2
13. 5
12. 0
14. 7
15. 7
10. 1
27. 2
2. 1
13. 1
9.7
13. 3
9.9
4. 3
5.9
6.1
3.0
8. 5
19.5
13. 3
28. 3
11. 8
AAY
16.9
2. 7
2.9
14. 6
2. 7
17.4
18. 8
0. 1
1. 3
4.4
5.4
0.9
0. 1
6.0
0. 1
0. 1
0. 2
3. 2
1. 1
18. 4
7. 5
Indicator acronyms: SECCHI = Secchi disc transparency (m)a
ISEC = inverse of Secchi disc transparency (m )a
COND = 'conductivity (micromhos cm~l)a
TPHOS = total phosphorus (mg 1-1)9
TON = total organic nitrogen (mg l~l)b
CHLA = chlorophyll _a (fxg I'1) a
AAY = algal assay control yield (mg dry wt) b
Lakes 1-38 are located in Minnesota; 39-68 are in Wisconsin; 69-86
are found in Michigan; 87-100 are in New York. The serial numbers
are unique to this report.
Mean values based on three sampling periods.
Values based on composite fall sample.
153
-------
APPENDIX B
SAMPLING DATES FOR 100 NES-SAMPLED LAKES (1972)
en
Lake Serial 1st
Name Number Sample
Blackduck
Bemidji
Andrusia
Wolf
Cass
Leech
Birch
Trout
Mashkenode
Whitewater
Pelican
Shagawa
Gull
Rabbit
Cranberry
Darling
Carlos
Le Homme
Dieu
Minnewaska
Nest
Green
Wagonga
Clearwater
Mud (at
Maple)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
07/12
07/11
07/11
07/11
07/11
07/11
07/12
07/11
07/10
07/06
07/10
07/08
07/02
07/02
07/02
07/06
07/10
07/06
07/06
07/02
07/02
07/02
07/03
06/30
2nd
Sample
09/08
09/08
09/08
09/08
09/07
09/08
09/03
09/08
09/09
09/02
09/07
09/07
09/05
09/04
09/04
09/01
09/02
09/02
09/01
08/31
08/31
08/31
08/29
08/29
3rd
Sample
10/21
10/21
10/21
10/21
10/21
10/21
10/24
10/22
10/19
10/25
10/22
10/22
10/24
10/24
10/24
10/25
10/28
10/28
10/25
10/25
10/25
10/25
10/27
10/26
Lake Serial
Name Number
Cokato
Buffalo
Carrigan
Silver
Minn e tonka
Forest
White Bear
St. Croix
Spring
Pepin
Madison
Sakatah
Bear
Albert Lea
Yellow
Wapogasset
Long
Elk
Trout
Crystal
Tainter
Shaw an o
Poygan
Butte
des Morts
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
1st
Sample
06/30
06/30
06/30
07/03
06/29
06/29
06/29
06/28
06/28
06/28
07/01
07/01
07/01
07/01
06/26
06/26
06/25
06/25
06/25
06/25
06/26
06/22
06/22
06/22
2nd
Sample
08/29
08/29
08/29
08/29
09/05
08/27
08/27
08/26
08/26
09/03
08/30
08/30
08/30
08/30
08/27
08/26
08/25
08/25
08/23
08/23
08/26
08/24
08/21
08/20
3rd
Sample
10/26
10/26
10/26
10/26
10/29
11/05
11/05
11/04
11/04
11/04
10/29
10/29
10/29
10/29
11/03
11/03
11/04
11/04
11/04
11/04
11/03
11/08
11/08
11/09
-------
en
APPENDIX B (continued)
SAMPLING DATES FOR 100 NES-SAMPLED LAKES (1972)
Lake Serial
Name Number
Winnebago
Round
Green
Swan
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac La Belle
Oconomowoc
Okauch.ee
Pine
Nagawicka
Pewaukee
Tichigan
BY owns
Middle
Delavan
Como
Geneva
Charlevoix
Higgins
Houghton
Pere
Marquette
White
Muskegon
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
1st
Sample
06/24
06/23
06/22
06/22
06/20
06/22
06/23
06/22
06/23
06/21
06/21
06/21
06/21
06/21
06/21
06/21
06/22
06/23
06/21
06/21
06/16
06/15
06/15
06/17
06/13
06/13
2nd
Sample
08/20
08/22
08/21
08/20
08/21
08/20
08/20
08/17
08/19
08/19
08/19
08/19
08719
08/19
08/17
08/16
08/19
08/17
08/16
08/16
09/14
09/16
09/20
09/18
09/18
09/19
3rd
Sample
11/09
11/08
11/08
11/10
11/09
11/10
11/10
11/10
11/09
11/11
11/10
11/09
11/10
11/10
11/10
11/10
11/10
11/10
11/10
11/10
11/12
11/12
11/14
11/13
11/14
11/14
Lake Serial
Name Number
Fremont
Mona
Crystal
Jordan
Thornapple
Strawberry
Chemung
Thompson
Ford
Union
Long
Randall
Schroon
Black
Cassadaga
Chatauqua
Conesus
Canandaigua
Keuka
Seneca
Cayuga
Owasco
Cross
Otter
Round
Saratoga 1
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
00
1st
Sample
06/13
06/13
06/15
06/15
06/13
06/17
06/16
06/16
06/16
06/14
06/13
06/14
06/01
05/20
05/26
05/26
05/27
05/27
05/27
05/16
05/16
05/28
05/28
05/17
05/28
05/15
2nd
Sample
09/15
09/19
09/17
09/18
09/18
09/19
09/19
09/19
09/19
09/16
09/17
09/16
07/25
07/25
07/27
07/27
07/27
07/21
07/21
07/23
07/23
07/24
07/24
07/24
07/25
07/25
3rd
Sample
11/13
11/14
11/14
11/15
11/14
11/13
11/15
11/15
11/13
11/12
11/12
11/12
10/10
10/10
10/13
10/12
10/13
10/14
10/14
10/14
10/13
10/12
10/13
10/13
10/10
10/11
-------
C71
cr>
APPENDIX C
MORPHOMETRY AND HYDROLOGY OF STUDY LAKES3
Lake Serial Area Mean Depth Maximum
Name Number (ha) (m) Depth (m)
Blackduck
Bemidji
Andrusia
Wolf
Cass
Leech
Birch
Trout
Mashkenode
Whitewater
Pelican
Shagawa
Gull
Rabbit
Cranberry
Darling
Carlos
Le Homme Dieu
Minnewaska
Nest
Green
Wagonga
Clearwater
Mud (at Maple
Lake)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
1 , 110
2,598
614
425
6, 312
45, 326
519
765
41
490
4,429
955
3, 812
340
8. 1
386
1, 020
706
2, 877
382
2, 187
654
1, 288
4.6
9.6
7.9
8. 5
7.6
4. 7
3. 1
15.2
2.1
6.1
3.7
5. 7
9.1
3. 7
3.4
6.2
13. 1
6.4
6.0
4.6
6.4
1.2
5. 2
8. 5
23. 2
18. 3
17. 7
36.6
45. 7
13. 7
41. 2
4. 3
22. 3
11.6
14. 7
26. 2
12. 8
10.4
18.9
49.7
25.9
9.8
12. 2
33. 5
4.6
22.9
Volume
(m3) x 106
50. 733
253.407
48.426
36. 299
480.934
214. 132
15. 826
116. 564
0. 872
29.701
162.005
54. 436
348. 508
43. 520
0. 275
23. 888
133. 660
45. 175
171. 893
17. 485
140. 032
7.978
66. 724
Shore Line
(km)
16. 0
23. 8
14. 7
12. 1
63. 1
22. 7
21. 7
3.06
21. 1
56. 1
29.0
64. 2
9. 81
10. 8
19.2
15. 3
29.6
19. 8
20.9
27.9
Retention
Timek
4. 3 y
268 d
47 d
37 d
313 d
5. 2y
17. 4y
9 d
5.0 y
0.9 y
2. 9 y
3.7y
7.9 y
12. 7 y
190 d
3. 7y
1.3y
1.4y
-------
en
--j
APPENDIX C (continued)
MORPHOMETRY AND HYDROLOGY OF STUDY LAKES
Lake Serial Area Mean Depth Maximum Volume
Name Number (ha) (m) Depth (m) (m ) x 10°
Cokato
Buffalo
Carrigan
Silver
Minnetonka
Forest
White Bear
St. Croix
Spring
Pepin
Madison
Sakatah
Bear
Albert Lea
Yellow
Wapogasset
Long
Elk
Trout
Crystal
Tainter
Shawano
Poygan
Butte des Morts
Winnebago
Round
Green
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
5,
1,
3,
2,
10,
1,
2,
4,
3,
55,
2,
220
611
66
170
855
893
077
322
392
118
541
497
556
985
926
480
169
36
544
36
709
491
449
584
730
5. 7
972
7.6
4. 2
0.9
1. 2
4. 3
3.4
6.9
8. 7
5. 1
4.0
2. 1
1. 1
5. 8
5.2
3.4
1. 8
11. 7
9.8
4. 0
3. 2
2. 1
1. 8
4.0
5. 7
31. 7
15. 2
9. 1
2.4
2. 1
27. 8
11. 3
24.4
23. 8
3. 1
17. 1
18. 0
3. 7
1. 1
1. 8
9. 8
9.8
16. 5
6.4
35. 1
21. 0
11. 3
10. 7
3. 4
2. 7
6.4
17. 1
71. 9
16
27
0
2
401
30
74
291
514
17
7.
10
53
25
5
0
180
3
27
80
94
65
2,208
0
939
. 775
. 007
. 594
. 040
. 727
. 204
. 152
.619
.979
. 847
607
. 591
. 330
. 248
. 632
.691
. 276
. 552
. 806
. 231
. 305
. 550
. 184
. 321
. 021
Shore Line
(km)
7.
9.
6.
175.
23.
21.
33.
15.
36.
11.
30.
2.
26.
116.
34.
15
25
87
0
1
6
8
8
0
3
6
25
6
9
1
Retention
Time
413
1. 4
3. 5
15
6.5
23
9
3. 3
44
73
194
15
2
C
45
1. 5
13
6
210
20. 7
d
y
y
y
y
d
d
y
d
d
d
d
d
d
y
d
d
d
y
-------
APPENDIX C (continued)
MORPHOMETRY AND HYDROLOGY OF STUDY LAKES
un
00
Lake Serial Area
Name Number (ha)
Swan
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac la Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Charlevoix
Higgins
Houghton
Pere Marquette
White
Muskegon
Fremont
Mona
Crystal
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
164
2, 671
1,099
555
4, 241
452
318
451
285
415
1,009
450
160
72
718
383
2, 130
6,985
3, 885
8, 112
224
1,040
1, 680
320
281
293
Mean Depth Maximum
(m) Depth (m)
9.8
1.4
5. 2
4.9
1. 5
4. 7
6.3
9. 5
11. 8
10.1
3. 1
1. 8
2.4
4.6
7.6
1. 2
18.6
16. 8
14.9
2. 3
6.7
7.0
10. 1
4.0
25.0
3.4
8. 8
17. 1
2. 1
14.0
18.9
28. 7
25.9
27.4
13. 7
19.2
13.4
13. 7
17. 1
2. 7
41.2
37. 2
41.2
6.4
11.6
21. 3
21.0
26. 8
12. 8
21. 3
Volume
(m3) x 106
15.908
37. 840
58. 234
26. 325
68. 829
21. 080
20.043
40. 320
33.448
41. 147
45.698
8. 096
3. 867
3. 293
54. 810
4. 795
395.928
1,170.944
580.230
187.901
71.468
120.295
32.157
11.487
Shore Line
(km)
63.7
15.4
14. 2
48.9
14.0
11. 3
24. 1
11. 8
13. 8
22. 1
20. 0
8.04
8. 53
28. 5
13.5
32. 5
54. 1
49.6
20. 8
18. 8
8.66
18.6
Retention
Time
179 d
155 'd
139 d
3. 6 y
24 d
8 m
10 m
1. 5y
4. 2 y
19 d
2. 8y
1. 6y
29. 9 y
3. 2 y
15. 6y
1.3y
56 d
23 d
1.9y
76 d
-------
un
APPENDIX C (continued)
MDRPHOMETRY AND HYDROLOGY OF STUDY LAKES
Lake
Name
Jordan
Thornapple
Strawberry
Chemung
Thompson
Ford
Union
Long
Randall
Schroon
Black
Cassadaga
Chautauqua
Conesus
Canandaigua
Keuka
Seneca
Cayuga
Owasco
Cross
Otter
Round
Saratoga
Winona
Trace
Calhoun
Serial Area
Number (ha)
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
174
166
104
126
106
425
213
85
208
1,671
3, 380
85
5, 720
1, 347
4, 300
4, 740
17, 252
17, 198
2, 746
881
114
127
1, 632
73
170
Mean Depth
(m)
7.
4.
6.
8.
2.
4.
0.
5.
5.
14.
2.
3.
6.
11.
39.
22.
88.
52.
29.
5.
7.
1.
8.
3
3
7
5
7
4
9
2
5
3
7
1
7
5
0
4
4
4
3
5
9
2
2
Maximum Volume
Depth(m) (m3) x 106
17. 7
9.5
15. 2
21. 3
15. 8
11.9
4. 3
12. 5
14.9
46. 3
5.2
8. 8
23. 5
18. 0
83. 5
56. 7
188.4
132. 6
54. 0
16. 8
29. 3
2.4
25.0
12.
7.
7.
10.
2.
18.
1.
4.
11.
239.
92.
2.
392.
154.
1,677.
1,063.
15,249.
9,375.
802.
49.
129.
0.
13.
995
063
006
821
909
521
943
425
263
315
718
603
266
760
377
356
165
141
997
339
308
876
940
Shore Line
(km)
7.
6.
7_
7.
23.
.5.
39.
92.
8.
68.
29.
57.
94.
121.
136.
39.
20.
5.
4.
37.
2.
19
34
93
35
2
89
9
7
21
3
7
8
0
3
5
8
0
86
35
1
74
Retention
Time
304
11
13.
4.
152
15.
2
31
41
153
0.
4
1..
2.
15
7.
33.
10.
7
150
1.
d
d
2 d
2 y
d
2 d
d
d
d
d
i y
m
4 y
3 y
y
8 y
i y
9 y
d
d
3 y
-------
CTl
O
APPENDIX C (continued)
MORPHOMETRY AND HYDROLOGY OF STUDY LAKES
Lake
Name
Big Stone
Zumbro
Oneida
Canadarago
Mendota
Monona
Waubesa
Cottonwood
Maple
Serial
Number
104
105
106
107
108
109
110
111
241
Area
(ha)
5, 103
345
20, 721
3,983
1, 350
855
150
240
Mean Depth
(m)
3. 4
1.6
5. 8
Maximum
Depth (m)
4.9
16. 7
2.4
23. 2
Volume Shore Line
(m3) x 106 (km)
17.35 96.4
0.240 5.63
13.920
Retention
Time
1.6 y
347 d
a This table has been compiled from information contained in the files of the Eutrophication Survey
Branch, Pacific Northwest Environmental Research Laboratory (NERC-Corvallis).
y = year, m = month, d = day
c Seepage lake.
-------
APPENDIX D
LANDSAT-1 MSS MODELS, CONCATENATIONS,
AREAL RELATIONSHIPS AND DESCRIPTIVE STATISTICS
This appendix contains information relevant to each LANDSAT-1 MSS Frame
examined during the investigation. It is divided into a series of
subappendices, one for each frame or pair of frames. Frames in juxta-
position on the same flightline are treated as one. Each subappendix
is in turn divided into five sections:
Section D_. 1. Regression Models and Correlation Coefficients
Section D_.2. Three-dimensional Color Ratio Model
Section D_.3. Concatenation of Extracted Lakes
Section D_.4. MSS-Lake Surface Area Relationships
Section D_.5. Lake MSS Descriptive Statistics
161
-------
APPENDIX Dl. 9 August 1972 (1017-16091, 1017-16093)
Section D1.1. Regression Models and Correlation Coefficients
The regression models for the prediction of Secchi disc transparency and
chlorophyll a^are in Section VI along with the coefficients of corre-
lation between the MSS data (colors and color ratios), the trophic
indicators and the multivariate trophic state index. The model for the
prediction of trophic state is in Section VII.
Section D1.2. Three-dimensional Color Ratio Model
The color ratio model is in Section VII (Figure 33).
Section D1.3. Concatenation of Extracted Lakes
The concatenation of lakes extracted from Frame 1017-16093 is in Section
V; the Frame 1017-16091 concatenation is in this appendix (Figure Dl-1).
Section Dl .4. MSS-Lake Surface Area Relationships
The estimates of lake surface area using MSS data are in Section VI
(Table 16).
Section Dl.5. Lake MSS Descriptive Statistics
The MSS statistics for Frame 1017-16093 are in Table Dl-1 and those for
Frame 1017-16091 are in Table Dl-2.
162
-------
P o
But t e Dei Mo r I
1017-16091
LAKES
IP£ 09AUG72 UIICDrr.IN c> EPA tiEI
:uri HPF 14. 1974 07152 3 JFL IPL
Figure Dl-1.
IR2 concatenation of five Wisconsin lakes extracted from
Frame 1017-16091 (9 August 1972). The images have not
been skewed to correct for geometric distortions. Cloud
cover is evident over the northern (upper) end of Lake
Winnebago.
163
-------
TABLE Dl-1. MSS DESCRIPTIVE STATISTICS FOR 15 WISCONSIN LAKES
EXTRACTED FROM FRAME 1017-16093 (9 AUGUST 1972)
Lake Name
Kegonsa
Rock
Koshkonong
Lac La Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
Pi xel
Count
2,675
1,009
9,247
944
652
871
568
751
1,996
689
309
117
1,541
703
4,543
Green
50.18a
1.63b
38.86
3.27
46.85
1.81
43.30
4.81
39.87
4.90
36.78
1.73
36.54
2.89
42.46
2.78
37.35
2.54
40.78
3.06
36.23
2.59
36.49
2.77
45.42
2.62
47.86
5.20
39.42
2.12
LANDSAT-1
Red
30.08
1.50
19.39
3.07
31.15
2.45
24.27
4.07
19.99
4.27
19.70
2.17
16.99
2.70
23.85
1.83
20.75
2.34
25.56
1.67
19.06
2.87
18.54
3.27
25.39
1.66
30.27
4.24
18.53
2.16
MSS Bands
IR1
24.38
3.30
13.62
3.38
27.95
2.93
15.11
3.35
14.46
4.48
14.35
4.27
13.04
3.95
17.06
3.71
16.19
4.44
23.21
4.28
15.78
3.99
17.38
6.53
22.07
3.00
22.51
3.25
13.71
2.81
IR2
8.34
5.24
5.10
3.74
10.08
2.89
5.69
3.73
6.35
4.72
.6.07
4.68
5.66
4.58
6.12
3.94
6.36
4.01
10.47
5.26
7.17
4.63
9.35
6.91
8.35
3.98
8.62
4.39
5.69
3.25
Mean DN value for the lake.
bStandard deviation for the lake pixels.
164
-------
TABLE Dl-2.
MSS DESCRIPTIVE STATISTICS FOR 5 WISCONSIN LAKES
EXTRACTED FROM FRAME 1017-16091 (9 AUGUST 1972)
Lake Name
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Serial
Number
47
48
49
51
53
Pixel
Count
9,177
7,395
114,186
6,613
5,162
LANDSAT-1 MSS Bands
Green
46.05a
4.82b
42.99
2.55
42.38
3.22
36.90
2.00
37.57
2.09
Red
29.50
5.07
26.74
1.83
24.95
2.67
17.03
2.32
22.76
2.00
IR1
23.62
5.25
23.08
2.78
20.86
5.50
12.12
3.21
20.12
4.24
IR2
9.55
5.24
8.25
3.59
6.65
3.92
4.59
3.18
7.48
4.57
aMean DN value for the lake.
bStandard deviation of the lake DN values.
165
-------
APPENDIX D2. 28 August 1972 (1036-16152)
Section D2.1. Regression Models and Correlation Coefficients
No models were developed because the sample size is insufficient; only
three NES-sampled lakes were extracted from the frame. The three lakes
(Rock, Kegonsa, and Koshkonong) have higher mean DN levels than on 9
August 1972, but have maintained their relative positions.
Section D2.2. Three-dimensional Color Ratio Model
No model was constructed.
Section D2.3. Concatenation of Extracted Lake
The concatenation of extracted lakes is displayed as Figure D2-1.
Section D2.4. MSS-Lake Surface Area Relationships
The estimates of lake surface using the MSS pixel counts and a conver-
sion factor of 0.48 (1 pixel = 0.48 hectares) are in Table D2-1.
Section D2.5. Lake MSS Descriptive Statistics
The MSS statistics are in Table D2-2.
166
-------
108
M P n d o ' o
109
4
5 5
Koch
o u b
5 4
9 o n i o
"
Figure D2-1.
IR2 concatenation of six Wisconsin lakes extracted from
Frame 1036-16152 (28 August 1972).
167
-------
TABLE D2-1. AREAL ASPECTS OF 6 WISCONSIN LAKES EXTRACTED FROM FRAME 1036-16152 (28 AUGUST 1972)
OD
Lake Name
Kegonsa
Rock
Koshkonong
Mendota
Monona
Waubesa
Serial
Number
54
55
56
108
109
110
Pixel
Count
2,805
1 ,039
9,364
8,460
2,729
1,938
LANDSAT-1 Lake
Area (ha)
l,340.8a
496. 6a
4,476.0a
4,043.9
1,304.5
926.4
Map Lake
Area (ha)
1 ,099.2
554.8
4,241.5
3,937.7
1,349.7
855.1
Map Area:LANDSAT-l
Area Ratio
0.820
1.117
0.948
0.974
1.035
0.923
aThe estimated areas of these lakes using the MSS pixel counts from Frame 1017-16093 are,
respectively: 1,277.4 hectares, 481.8 hectares, and 4,415.9 hectares.
-------
TABLE D2-2.
MSS DESCRIPTIVE STATISTICS FOR 6 WISCONSIN LAKES
EXTRACTED FROM FRAME 1036-16152 (28 AUGUST 1972)
Lake Name
Kegonsa
Rock
Koshkonong
Mendota
Monona
Waubesa
Serial
Number
54
55
56
108
109
110
Pixel
Count
2,805
1,039
9,364
8,460
2,729
1,938
Green
46.97*
2.01°
42.90
2.38
43.47
2.72
40.07
2.01
39.24
2.84
46.78
4.01
LANDSAT-1
Red
26.62
1.40
22.93
1.74
27.82
1.64
21.71
1.83
21.84
2.17
26.50
2.07
MSS Bands
IR1
17.88
2.64
15.22
3.92
22.23
2.55
13.88
2.97
14.29
3.38
20.28
3.08
IR2
6.57
3.41
6.78
4.49
8.19
3.15
5.24
3.31
6.05
3.70
7.92
4.03
aMean DN value for the lake.
^Standard deviation of the lake DN values,
169
-------
APPENDIX D3. 11 June 1973 (1323-16100, 1323-16094)
Section D3.1. Regression Models and Correlation Coefficients
The regression model for the prediction of trophic state is found in
Section VII; it is of little practical value.
Section 3.2. Three-dimensional Color Ratio Model
The color ratio model is in Section VII (Figure 34).
Section 3.3. Concatenations of Extracted Lakes
The 23 lakes extracted from the frames are displayed in the form of two
concatenations, Figures D3-1 and D3-2. The fragmented appearance of
Lake Geneva is a result of cloud cover. Lake Winnebago is partially
truncated, a consequence of separating the continuous MSS strip into
discrete frames. Cloud cover is responsible for the mottled appearance
of Lake Poygan. Middle Lake has not been excised from the Lauderdale
Lakes as was the case for 9 August 1972.
Section D3.4. MSS-Lake Surface Area Relationships
The estimates of lake surface area using the MSS pixel count are found
in Table D3-1. Area estimates have not been determined for truncated
water bodies or those partially covered by clouds.
Section D3.5. Lake MSS Descriptive Statistics
The MSS statistics are in Table D3-2.
170
-------
y»JHr^ 68
G e n e v
J^»»
5 9
Okouc hee
6 3
I i C. h t g Q n
P,n<
6 1
Nogow icko
6 5
Middle
(Lauderdole Loket)
6 6
Oe I
6 2
P ewaukee
58
O c o n om o w o c
5 7
locla Belle
6 7
:" :•'•
-
Figure D3-1.
IR2 concatenation of 12 Wisconsin lakes extracted from
Frame 1323-16100 (11 June 1973). The 11 other lakes
extracted from the frame are in Figure D3-2.
171
-------
Ko t h k on on g
IUUH::-: UI:CQM:IN WE:
Figure D3-2.
IR2 concatenation of 11 Wisconsin lakes extracted from
Frame 1323-16100 (11 June 1973). The other 12 lakes
extracted from the frame are in Figure D3-1.
172
-------
TABLE D3-1.
AREAL ASPECTS OF 23 WISCONSIN LAKES EXTRACTED FROM
FRAMES 1323-16094 and 1323-16100 (11 JUNE 1973)
CO
Lake Name
Poygan
Butte Des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac La Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Del a van
Como
Geneva
Mendota
Monona
Waubesa
Serial
Number
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
108
109
no
Pixel
Count
11,829
8,053
98,821
6,742
5,485
2,706
1,127
9,516
939
627
928
550
775
1,982
707
304
563
1,517
740
3,555
8,488
2,791
1,882
LANDSAT-1 Lake
Area (ha)
5,648.9
3,845.3
47,206.8
3,220.7
2,618.8
1,292.7
538.4
4,545.8
448.6
299.5
443.3
262.7
370.2
945.8
337.7
145.2
268.9
724.7
353.5
1,698.2
4,054.7
1 ,333.3
889.0
Map Lake
Area (ha)
4,448.5
3,584.4
55,730.4
2,972.9
2,671.0
1,099.2
554.8
4,241.3
452.1
317.7
450.8
284.5
415.2
1,008.9
449.8
160.3
337.6
717.9
383.0
2,129.5
3,937.7
1,349.7
855.1
Map Area:LANDSAT-1
Area Ratio
a
0.932
1.181b
0.923
1.010
0.850
1.031
0.933
1.008
1.061
1.017
1,083
1.122
1.066
1.332
1.104
1.254C
0.991
1.084,
d
0.971
1.012
0.951
aThe extracted image includes Lake Winneconne and substantial cloud cover.
^The northern end of the lake is outside the sensor field of view.
cThe Lauderdale lakes complex was extracted and treated as a single lake.
dThe presence of cloud cover has resulted in the low LANDSAT-1 area estimate,
-------
TABLE D3-2. MSS DESCRIPTIVE STATISTICS FOR 23 WISCONSIN LAKES
EXTRACTED FROM FRAME 1323-16100 (11 JUNE 1973)
Lake Name
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac La Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Mendota
Monona
Waubesa
Serial
Number
47
48
49
51
53
54
55
57
57
58
59
60
61
62
63
64
65
66
67
68
108
109
110
Pixel
Count
11,829
8,053
98,821
6,742
5,485
2,706
1,127
9,516
939
627
928
550
775
1,982
707
304
563
1,517
740
3,555
8,488
2,791
1,882
LANDSAT-1 MSS Bands
Green
50.53
4.97
47.28
2.83
49.43
2.98
43.53
2.59
51.55
1.99
49.01
2.48
50.02
3.07
51.94
2.40
52.02
4.28
46.24
3.68
49.49
2.49
51.83
2.45
46.71
2.65
53.19
5.65
48.19
2.82
50.40
2.49
48.94
3.04
51.76
3.05
60.33
5.56
54.15
5.71
51.01
2.58
49.27
3.08
47.56
2.76
Red
33.74
5.24
30.78
2.73
32.87
3.01
23.59
2.67
34.55
2.57
29.20
2.03
29.41
2.21
36.81
3.31
32.55
4.34
26.80
3.15
31.32
2.28
27.97
2.20
27.70
2.18
32.02
3.70
31.03
2.37
29.75
2.37
28.25
2,46
31.16
2.72
40.42
5.37
31.44
4.87
28.74
2.36
27.68
2.39
27.53
2.16
IR1
25.52
5.89
22.83
4.06
20.14
2.65
16.93
3.67
27.40
3.34
19.67
3.50
22.14
3.35
25.61
2.97
22.49
3.49
20.38
3.72
22.68
4.13
21.03
3.47
23.09
5.17
24.20
4.24
24.85
5.11
22.03
3.84
22.74
5.31
24.30
5.01
28.28
4.02
23.63
5.23
19.80
3.19
19.81
4.39
21.31
6.56
IR2
13.95
5.69
11.47
4.14
8.80
2.80
7.75
3.75
13.25
3.75
9.18
3.49
11.66
3.73
11.26
3.43
11.37
3.92
10.65
4.09
11.75
4.69
10.99
4.20
11.77
4.74
12.08
4.15
13.54
5.04
12.11
4.31
13.05
5.22
12.29
4.10
14.07
4.22
13.26
4.90
8.59
3.16
9.36
3.90
10.38
5.28
174
-------
APPENDIX D4. 17 July 1973 (1359-16091, 1359-16094)
Section D4.1. Regression Models and Correlation Coefficients
The regression model for the prediction of the multivariate trophic
state index is in Section VII.
Section D4.2. Three-dimensional Color Ratio Model
The MSS color ratio model is in Section VII (Figure 35). It is very
similar in appearance to the model constructed from 9 August 1972 data.
Section D4.3. Concatenations of Extracted Lakes
Eighteen of the 21 lakes extracted from the frames are found in Figures
D4-1 and D4-2. Lake Winnebago is common to both frames. Cloud inter-
ference is noted over some of the lakes.
Section D4.4. MSS-Lake Surface Area Relationships
The estimates of lake surface area using the MSS pixel counts are found
in Table D4-1.
Section D4.5. Lake MSS Descriptive Statistics
The MSS statistics for Frame 1359-16091 are in Table D4-2 and those for
Frame 1359-16094 are in Table D4-3.
175
-------
63
T i c h i g a n
64
f o w n s
^62
,65
Middle
(louderdole Iokei)
W67
C o mo
" i n n e b 090
1359-16094 IRS 17JUL73 UISCDHSIH
WE:
BISECT
Figure D4-1.
IR2 concatentation of 15 Wisconsin lakes extracted from
Frame 1359-16094 (17 July 1973).
176
-------
4 8
Butte Des Morts
1259-lt.ri91 IP£ 17JUL73 E-HE WISCONSIN LAKES
INSECT
Figure D4-2.
IR2 concatenation of 4 Wisconsin lakes extracted from
Frame 1359-16091 (17 July 1973).
177
-------
TABLE D4-1. AREAL ASPECTS OF 21 WISCONSIN LAKES EXTRACTED FROM
FRAMES 1359-16091 and 1359-16094 (17 JULY 1973)
"-J
00
Lake Name
Shawano
Poygan
Butte des Morts
Winnebago
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac La Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
46
47
48
49
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
Pixel
Count
5,263
15,836
7,600
112,887
6,594
5,057
2,716
1,102
9,029
927
613
791
563
505
1,851
561
317
630
1,487
615
3,684
LANDSAT-1 Lake
Area (ha)
2,514.1
7,564.9
3,630.5
53,926.1
3,150.0
2,415.7
1,297.4
526.4
4,313.2
442.8
292.8
377.9
269.0
241.2
884.2
268.0
151.4
301.0
710.3
293.8
1,759.9
Map Lake
Area (ha)
2,491.3
4,448.5
3,584.4
55,730.4
2,972.9
2,671.0
1,099.2
554.8
4,241.3
452.1
317.7
450.8
284.5
415.2
1,008.9
449.8
160.3
337.6
717.9
383.0
2,129.5
Map Area:LANDSAT-l
Area Ratio
0.991
a
0.987
1.034
0.944
1.106
0.847
1.054
0.983
1.021
1.085
1.193
1.058,
_b
1.141b
c
1.059J
1.122d
1.011.
1.304?
1.210°
aPixel count includes Lake Winneconne.
bCloud interference.
CA portion of the lake body was truncated during the extraction process.
Includes the entire Lauderdale lake complex.
-------
TABLE D4-2. MSS DESCRIPTIVE STATISTICS FOR 4 WISCONSIN LAKES
EXTRACTED FROM FRAME 1359-16901 (17 JULY 1973)
Lake Name
Poygan
Butte des Morts
Winnebago
Shawano
Serial
Number
47
48
49
46
Pixel
Count
15,836
7,600
112,887a
5,263
LANDSAT-1 MSS Bands
Green
56.74
2.96
53.31
3.04
51.41
3.29
46.38
2.69
Red
39.27
3.45
34.49
3.45
32.40
3.31
28.38
2.89
IR1
32.09
4.37
28.99
4.51
23.94
3.62
20.61
3.57
IR2
16.74
3.32
13.44
3.67
10.74
3.35
10.16
3.54
aSouthern end of Lake Winnebago is outside the sensor field of view.
179
-------
TABLE D4-3. MSS DESCRIPTIVE STATISTICS FOR 17 WISCONSIN LAKES
EXTRACTED FROM FRAME 1359-16094 (17 JULY 1973)
Lake Name
Green
Beaver Dam
Kegonsa
Rock
Koshkonong
Lac La Belle
Oconomowoc
Okauchee
Pine
Nagawicka
Pewaukee
Tichigan
Browns
Middle
Delavan
Como
Geneva
Serial
Number
51
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
Pixel
Count
6,594
5,057
2,716
1,102
9,029
927
613
791
563
505
1,851
561
317
630
1,487
615
3,684
LANDSAT-1 MSS Bands
Green
52.65
2.27
56.88
2.40
60.79
2.59
51.85
3.04
54.72
2.79
56.55
5.43
49.58
3.45
50.77
3.57
53.20
4.07
51.92
6.11
52.27
3.75
52.04
4.75
53.91
2.12
46.91
2.95
57.85
3.49
59.61
4.50
52.22
4.11
Red
29.40
2.10
38.74
2.72
36.33
2.25
30.38
3.13
37.93
3.52
34.60
5.24
28.35
3.31
30.91
3.46
28.25
3.37
31.30
5.26
31.26
2.88
33.00
3.60
31.25
2.28
25.95
2.71
33.98
2.87
40.60
4.35
28.12
3.68
IR1
21.00
3.37
33.17
3.80
20.71
4.06
21.99
4.44
29.03
3.89
23.29
4.48
21.04
3.86
23.42
5.13
22.35
4.87
25.04
5.73
24.50
5.26
28.48
5.43
24.09
3.39
22.87
5.07
30.45
4.78
28.89
4.43
22.14
4.57
IR2
10.69
3.35
14.70
3.40
13.97
3.83
11.42
4.37
12.26
3.21
12.00
4.64
10.97
4.17
12.99
5.27
12.95
5.51
13.91
5.46
13.04
4.41
15.30
5.08
13.07
3.72
10.73
6.94
14.74
4.03
15.20
4.35
12.67
4.71
180
-------
APPENDIX D5. 14 August 1972 (1022-16373)
Section D5.1.
Regression Models
[Table D5-1)
and Correlation Coefficients
Twelve Minnesota lakes were extracted from the frame; Maple Lake
(241) was not included in the final regression model efforts.
Darling (16)
Le Homme Dieu (18)
Minnewaska (19)
Nest (20)
Green (21)
Wagonga (22)
Clearwater (23)
Maple (241)
Cokato (25)
Buffalo (26)
Silver (28)
Minnetonka (29)
TABLE D5-1. CORRELATIONS BETWEEN GROUND TRUTH AND MSS DATA (COLORS
AND COLOR RATIOS) FOR 11 MINNESOTA LAKES IN FRAME
1022-16373 (14 AUGUST 1972).
PCI
CHLA
LNCHLA
SECCHI
LNSECCHI
GRN
RED
IR1
IR2
GRNRED
GRNIR1
GRNIR2
REDIR1
REDIR2
IR1IR2
0.422
0.547
0.890
0.764
-0.585
-0.920
-0.644
-0.940
-0.636
0.536
0.487
0.613
0.763
0.758
-0.646
-0.704
-0.602
-0.670
-0.568
0.195
0.461
0.610
0.620
0.698
-0.725
-0.539
-0.558
-0.454
-0.495
-0.052
-0.749
-0.766
-0.791
-0.626
0.415
0.557
0.268
0.562
0.220
-0.597
-0.734
-0.801
-0.920
-0.801
0.492
0.728
0.475
0.728
0.439
-0.518
The best regression model for the prediction of trophic state is
A
PCI = 13.150 - 10.626 GRNIR1 + 2.327 GRNIR2
It explains about 96 percent of the variation about the mean (Table D5-2)
The observed and predicted PCI values are in Table D5-3.
181
-------
TABLE D5-2. ANALYSIS OF VARIANCE TABLE OF PCI REGRESSION MODEL FOR
11 MINNESOTA LAKES IN FRAME 1022-16373 (14 AUGUST 1972),
Source
Analysts of Variance
df Sum of Squares Mean Square Calculated F
Total (corrected) 10
42.423
4.242
Regression
Residual
2
8
40.750
1.673
20.375
0.209
97.488
R2 = 0.9606 x 100 = 96.06%
s.e. of estimate = 0.456
TABLE D5-3. PCI RESIDUALS OF 11 MINNESOTA LAKES IN FRAME 1022-16373
(14 AUGUST 1972).
Lake Name
Darling
Le Homme Dieu
Minnewaska
Nest
Green
Wagonga
Clearwater
Cokato
Buffalo
Silver
Minnetonka
Serial
Number
16
18
19
20
21
22
23
25
26
28
29
PCI
-0.73
-1.06
-0.32
0.77
-1.10
4.40
0.01
1.61
2.31
4.79
0.73
PCI
-0.34
-0.84
-0.68
1.04
-1.00
4.65
0.08
1.05
2.92
4.23
0.29
PC1-PC1
-0.39
-0.22
0.36
-0.27
-0.10
-0.25
-0.07
0.56
-0.61
0.46
0.44
The best model for the prediction of Secchi disc transparency is:
LNSECCHI = -4.105 - 154.13 GRNIR1 + 76.641 GRNIR2 + 265.290
REDIR1 - 130.200 REDIR2
It explains about 87 percent of the variation about the mean (Table D5-4),
The observed and predicted Secchi disc transparency values are in Table
D5-5. Caution is advised in using this model because a relatively large
number of variables are incorporated into it.
182
-------
TABLE D5-4. ANALYSIS OF VARIANCE TABLE OF THE SECCHI DISC TRANSPARENCY
REGRESSION MODEL FOR 11 MINNESOTA LAKES IN FRAME 1022-
16373 (H AUGUST 1972).
Analysis of Variance
Source
df Sum of Squares Mean Square Calculated F
Total (corrected) 10
9.529
0.953
Regression
Residual
4
6
8.324
1.205
2.081
0.201
10.353
R2 = 0.8735 x 100 = 87.35%
s. e. of estimate - 0.448
TABLE D5-5. SECCHI DISC TRANSPARENCY RESIDUALS OF 11 MINNESOTA LAKES
IN FRAME 1022-16373 (14 AUGUST 1972).
Lake Name
Darling
Le Homme Dieu
Minnewaska
Nest
Green
Wagonga
Clearwater
Cokato
Buffalo
Silver
Minnetonka
Serial
Number
16
18
19
20
21
22
23
25
26
28
29
SECCHI
2.79
1.73
1.55
0.97
2.29
0.23
1.80
2.74
1.98
0.15
1.47
SECCHI
1.87
2.45
1.73
0.97
2.68
0.16
1.25
2.82
1.44
0.31
1.60
SECCHI-S"ECCHI
0.92
-0.72
-0.18
0.00
-0.39
0.07
0.55
-0.08
0.54
-0.16
-0.13
Efforts to construct a regression model for the prediction of chlorophyll
a^ drew negative results.
183
-------
Section D5.2. Three-dimensional Color Ratio Model
The color ratio model ts displayed in Figure D5-1. Wagonga Lake and
Stiver Lake are isolated from the other lakes because their "IR1 values
exceed thetr RED DN values. The two lakes are often referred to as
being hypereutrophtc.
Sectton D5.3. Concatenation of Extracted Lakes
The Frame 1022-16373 concatenation is fn Figure D5-2. Portions of Lake
Minnetonka and Lake Le Homme Dieu were outside the sensor field of view,
resulting in the linear shore line effects. Minnetonka is a complex
lake consisting of 15 large bays.
Section D5.4. MSS-Lake Surface Area Relationships
The areal aspects of Frame 1022-16373 are in Table D5-6.
Section D5.5. Lake MSS Descriptive Statistics
The MSS statistics for Frame 1022-16373 are in Table D5-7.
184
-------
CO
en
-1.06
-0.73
©Darling
1 g| | cleafwater
(35) Cokato I 0.01
1 (S) Nest 1
4.4O 1 1
~^^ W agonga 1 1 1
©Maple
4.79
(^1) Silver 1 11 I
2 .31 | 1
/-^ Buffalo
T
1
Le Homme Dieu
Minnetonka
' 0.73
-0.3Z eva ,
^r
i M ' n ii c
G reen fa)
-1.10 T
Figure D5-1. Three-dimensional MSS color ratio model of 12 Minnesota lakes extracted from Frame
1022-16373 (14 August 1972).
-------
16
Do ' I i n g
18
le Hornme Die <
70
Neil
2 1
Green
22
Wog 009 a
2 3
C I e o rw
241
Mople
2 5
C o ko to
26
Buflolc
28
S.I,
E-1637
ECT
14Hin MINHES UH
Figure D5-2.
IR2 concatenation of 12 Minnesota lakes extracted from
Frame 1022-16373 (14 August 1972).
186
-------
TABLE D5-6. AREAL ASPECTS OF 12 MINNESOTA LAKES EXTRACTED FROM FRAME 1022-16373 (14 AUGUST 1972).
00
Lake Name
Darling
Le Homme D.ieu
Minnewaska
Nest
Green
Wagonga
Clearwater
Maple
Cokato
Buffalo
Silver
Minnetonka
Serial
Number
16
18
19
20
21
22
23
241
25
26
28
29
Pixel
Count
824
1,127
6,996
778
4,846
1,339
2,561
518
470
1,334
367
9,787
LANDSAT-1 Lake
Area (ha)
393.5
538.2
3,340.9
376.3
2,314.2
639.4
1,223.0
247.4
224.5
637.1
175.3
4,673.7
Map Lake
Area (ha)
386.1
765.7
3,144.5
382.4
2,355.8
725.2
1.287.8
287.3
220.2
611.1
170.8
5,855.6
Map Area: LANDSAT-1 Area
Ratio
0.981
_a
0.941
1.016
1.018
1.134
1.053
1.161
0.981
0.959
0.974
a
JA portion of the lake was outside the sensor field of view.
-------
TABLE D5-7. MSS DESCRIPTIVE STATISTICS FOR 12 MINNESOTA LAKES EXTRACTED
FROM FRAME 1022-16373 (14 AUGUST 1972).
Lake Name
Darling
Le Honyne Dieu
Mirmewaska
Nest
Green
Wagonga
Clearwater
Maple
Cokato
Buffalo
Silver
Minnetonka
Serial
Number
16
18
19
20
21
22
23
241
25
26
28
29
Pixel
Count
825
1,163
6.996
788
4,846
1,339
2.561
5.8
470
1,334
376
9.787
LANDSAT-1 MSS Bands
Green
49.30!*
4.00b
51.76
3.01
50.56
2.65
46.81
2.67
46.25
2.06
55.77
3.48
47.64
2.47
44.22
2.16
41.37
2.87
46.68
2.28
55.75
2.46
53.81
5.30
Red
29.37
4.27
30.90
3.33
28.74
2.11
27.60
1.98
25.78
1.59
34.92
2.96
27.82
2.09
25.38
1.80
25.19
1.80
26.49
1.46
33.86
2.30
31.60
4.31
IR1
23.89
5.28
24.02
4.45
21.66
3.32
23.54
4.62
19.12
2.74
36.92
3.33
22.00
4.56
22.23
4.81
22.37
4.19
25.36
4.37
35.89
2.72
24.11
4.87
IR2
13.59
5.57
13.51
5.03
10.72
3.97
12.08
5.48
9.31
3.34
17.18
3.90
11.15
5.14
12.43
5.26
12.74
5.39
11.64
3.88
17.10
4.10
11.53
4.85
? Mean DN value for the lake.
Standard deviation of the lake DN values.
188
-------
APPENDIX D6. 6 OCTOBER 1973 0075-16321)
Section D6.1. Regression Models, and Correlation Coefficients
(Table D6-1)
Sixteen Minnesota lakes were extracted from the frame. Lakes Sakatah,
Calhoun, Zumbro, and Maple were not used in the construction of the
regression model.
Clearwater (23 Minnetonka (29) Madison (35)
Maple (241) Forest (30) Sakatah (36)
Cokato (25) White Bear (31) Calhoun (103)
Buffalo (26) St. Croix (32) Zumbro (105)
Carrigan (27) Spring (33)
Silver (28) Pepin (34)
Table D6-1. CORRELATIONS BETWEEN GROUND TRUTH AND MSS DATA
(COLOR RATIOS) FOR 12 MINNESOTA LAKES IN FRAME 1075-16321
(6 OCTOBER 1972).
PCI CHLA LNCHLA SECCHI LNSECCHI
GRNRED
GRNIR1
GRNIR2
REDIR1
REDIR2
IR1IR2
-0.375
-0.793
-0.620
-0.663
-0.393
0.220
0.067
-0.391
-0.425
-0.471
-0.410
-0.146
-0.197
-0.494
-0.305
-0.425
-0.173
0.205
0.582
0.706
0.424
0.425
0.105
-0.349
0.517
0.827
0.545
0.606
0.250
-0.353
The best model for the prediction of trophic state is:
PCI = 11.553 - 7.132 REDIR 1
This model explains about 44 percent of the variation about the mean
(Table D6-2). It is of little practical value. The observed and
predicted PCI values are in Table D6-3.
189
-------
TABLE D6-2. ANALYSIS OF VARIANCE TABLE FOR THE PCI REGRESSION MODEL
FOR 12 MINNESOTA LAKES EXTRACTED FROM FRAME 1075-16321
(6 OCTOBER 1972).
Source
Analysis of Variance
df Sum of Squares Mean Square Calculated F
Total [Corrected)
11
R2 = 0.4391 x 100 = 43.91%
42.274
3.843
Regression
Residual
1
10
18.561
23.713
18.561
23.713
7.827
s. e. of estimate = 1.540
TABLE D6-3. PCI RESIDUALS OF 12 MINNESOTA LAKES EXTRACTED FROM FRAME
1075-16321 (6 OCTOBER 1972).
Lake Name
Clear-water
Cokato
Buffalo
Carrigan
Silver
Minnetonka
Forest
White Bear
St. Croix
Spring
Pepin
Madison
Serial
Number
23
25
26
27
28
29
30
31
32
33
34
35
PCI
0.01
1.61
2.31
4.40
4.79
0.73
-1.22
-1.41
-0.17
2.33
2.10
1.36
PCI
0.95
0.84
0.14
3.11
4.81
1.06
0.45
1.31
1.40
1.22
0.87
0.67
PCI -PCI
-0.94
0.77
2.17
1.29
-0.02
-0.33
-1.67
-2.72
-1.57
1.10
1.23
0.69
Section D6.2. Three-dimensional Color Ratio Model
The color ratio model is displayed in Figure D6-1.
Section D6.3. Concatenation of Extracted Lakes
The 15 lakes are displayed in two concatenations, Figure D6-2 and
Figure D6-3. A portion of Lake Pepin is outside the sensor field of
190
-------
view, accounting for the linear shoreline. The lake images are in
scale and this results in the very small image of Lake Zumbro. Lake
St. Croix was truncated during processing.
Section D6.4. MSS-Lake Surface Area Relationships
The areal aspects of the lakes extracted from Frame 1075-16321
are in Table D6-4.
Section D6.5. Lake MSS Descriptive Statistics
The MSS statistics are in Table D6-5.
191
-------
2.33
(33) Spring
2.10
Z umbro
10
ro
4.79
Silver
4.41
Carngan
GR N ' I R 1
Figure D6-1.
Three-dimensional MSS color ratio model of 15 Minnesota lakes extracted from
Frame 1075-16321 (6 October 1972).
-------
2 41
Maple
27
Minnc t on k c
28
S.I,
2 5
C o k a t o
CO
103
C a I hour
23
C I eo f w a t i
2 6
-dole
1075-16321 IF'2 6Qi~J72 MINNESOTA LAKES
INSECT
Figure D6-2. IR2 concatenation of 8 Minnesota lakes extracted from Frame 1075-16321 (6 October 1972)
Figure D6-3 contains 8 additional lakes extracted from the same frame.
-------
'or.,,
Wh,., B«<
36
5 ok o t o h
3 3
Spring
Si C,
35
M od. ion
1075-16321 IPS
INSECT
C'UL I i C
MINNESOTA
LAKES
Figure D6-3.
IR2 concatenation of 8 Minnesota lakes extracted from
Frame 1075-16321 (6 October 1972).
194
-------
TABLE D6-4. AREAL ASPECTS OF 15 MINNESOTA LAKES EXTRACTED FROM FRAME 1075-16321 (6 OCTOBER 1972)
en
Lake Name
Clearwater
Maple
Cokato
Buffalo
Carrigan
Silver
Minnetonka
Forest
White Bear
St. Croix
Spring
Pepin
Madison
Calhoun
Zumbro
Serial
Number
23
241
25
26
27
28
29
30
31
32
33
34
35
103
105
Pixel
Count
2,797
561
481
1,350
125
386
12,131
1 ,945
2,233
2,202
5,338
16,910
1,224
363
129
LANDSAT-1 Lake
Area (ha)
1,333.7
268.0
229.8
644.9
59.7
184.4
5,795.0
929.1
1,066.7
1,051.9
2,550.0
8,077.9
584.7
173.4
61.6
Map Lake
Area (ha)
1,287.8
287.3
220.2
611.1
65.6
170.8
5,855.6
892.8
1,076.5
3,322.2
2,391.8
10,117.5
541.1
169.6
344.8
Map Area: LANDSAT-1
Area Ratio
0.966
1.072
0.958
0.948
1.099
0.926
1.011
0.961
1.009
a
0.938.
D
0.925
0.978
c
aThe entire image was not extracted from the CCT's.
bA portion of the lake was outside the sensor field of view.
cThe disparity may be related to the long sinuous shape of the water body.
-------
TABLE D6-5. MSS DESCRIPTIVE STATISTICS FOR 15 MINNESOTA LAKES
EXTRACTED FROM FRAME 1075-16321 (6 OCTOBER 1972)
Lake Name
Clearwater
Maple
Cokato
Buffalo
Carrigan
Silver
Mi nne tonka
Forest
White Bear
St. Croix
Spring
Pepin
Madison
Zumbro
Calhoun
Serial
Number
23
241
25
26
27
28
29
30
31
32
33
34
35
105
103
Pixel
Count
2,797
561
481
1,350
125
386
12,131
1,945
2,233
2,202
5,338
16,910
1,224
129
363
LANDSAT-1 MSS Bands
Green
33.133
2.42b
32.95
2.42
31.53
2.90
35.22
2.16
32.12
2.34
41.15
1.61
35.90
2.49
34.35
2.09
33.36
2.98
31.19
2.64
34.64
2.28
33.72
2.09
35.37
2.23
38.29
5.44
33.79
2.18
Red
17.03
2.93
16.49
2.81
16.31
2.72
20.43
1.95
16.71
2.75
23.20
1.64
18.61
2.54
18.76
2.39
16.63
3.60
16.84
3.46
22.52
2.47
21.56
2.27
19.50
2.35
27.39
2.83
17.15
3.12
IR1
11.46
5.70
11.59
6.11
10.86
5.66
12.77
4.25
14.12
6.40
24.56
2.95
12.65
5.23
12.05
4.84
11.57
6.25
11.83
4.98
15.55
4.76
14.39
4.40
12.78
5.24
23.05
5.20
11.32
5.28
IR2
5.14
5.97
5.16
4.46
7.11
8.83
5.09
4.99
5.20
5.17
6.60
5.88
5.67
14.17
4.93
.Mean DN value for the lake
Standard deviation of the lake DN values.
196
-------
APPENDIX D7. 8 October 1972 (1077-16431)
Section D7.1. Regression Models, and Correlation Coefficients
(Table D7-1)
Ten lakes were extracted from the frame; Lake Winona, Trace Lake, and
Big Stone Lake were not used to develop the regression model.
Darling (16)
Carlos (17)
Le Homme Dieu
(18)
Minnewaska (19)
Nest (20)
Green (21)
Wagonga (22)
Winona (101)
Trace (102)
Big Stone (104)
TABLE D7-1.
CORRELATIONS BETWEEN GROUND TRUTH AND MSS DATA
(COLOR RATIOS) FOR 7 MINNESOTA LAKES EXTRACTED
FROM FRAME 1077-16431 (8 OCTOBER 1972)
PCI
CHLA
LNCHLA
SECCHI LNSECCHI
GRNRED
GRNIR1
GRNIR2
REDIR1
REDIR2
IR1IR2
-0.975
-0.941
-0.770
-0.918
-0.675
0.453
-0.936
-0.962
-0.786
-0.963
-0.716
0.466
-0.945
-0.968
-0.918
-0.953
-0.864
0.168
0.876
0.738
0.474
0.679
0.370
-0.581
0.935
0.858
0.579
0.826
0.473
-0.660
The best model for the prediction of trophic state is:
PCI = 34.509 - 18.548 GRNRED
This model explains about 95 percent of the variation about the mean
(Table D7-2). The observed and predicted PCI values are in Table D7-3
197
-------
TABLE D7-2. ANALYSIS OF VARIANCE TABLE OF THE PCI REGRESSION
FOR 7 MINNESOTA LAKES EXTRACTED FROM FRAME 1077-
16431 (8 OCTOBER 1972)
Analysis of Variance
Source
df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
6
1
5
25.269
24.035
1,233
4.212
24.035
0.247
97.433
R2 = 0.9512 x 100 = 95.12% s.e. of estimate = 0.497
TABLE D7-3. PCI RESIDUALS OF 7 MINNESOTA LAKES EXTRACTED
FROM FRAME 1077-16431 (8 OCTOBER 1972)
Lake Name ?.erjal
Number
Darling
Carlos
Le Homme Dieu
Minnewaska
Nest
Green
Wagonga
16
17
18
19
20
21
22
PCI
-0.73
-1.55
-1.06
-0.32
0.77
-1.10
4.40
PCI
-0.92
-1.18
-0.68
-0.06
1.09
-1.90
4.07
PCl-pTl
0.19
-0.37
-0.38
-0.26
-0.32
0.80
0.33
Section D7.2. Three-dimensional Color Ratio Model
The MSS color ratio model is displayed in Figure D7-1. The PCI values
are in very good agreement with the model.
Section D7.3. Concatenation of Extracted Lakes
The 10 lakes are shown in Figure D7-2.
Section D7.4. MSS-Lake Surface Area Relationships
The area! aspects of the lakes extracted from Frame 1077-16431 are in
Table D7-4.
198
-------
Section D7.5. Lake MSS Descriptive Statistics
The MSS statistics are presented in Table D7-5.
-------
4.40
££ ^-.Wagonga
§ i©-^—
Big Stone
-0.32
O) Kflinnewaska
- 1.10
Green
G R N I R 1
Figure D7-1. Three-dimensional MSS color ratio model of 10 Minnesota lakes extracted from
Frame 1077-16431 (8 October 1972).
-------
20
Ne il
16
Dor I ing
21
Green
18
Le Homme Diet
101
W i n o n o
102
Trace
Mi n n e wo ska
22
Wagon go
1077-16431 IP£ 05QCT72 MINNESOTA LAKES
INSECT
Figure D7-2.
IR2 concatenation of 10 Minnesota lakes extracted from
Frame 1077-16431 (8 October 1972).
201
-------
TABLE D7-4. AREAL ASPECTS OF 10 MINNESOTA LAKES EXTRACTED FROM FRAME 1077-16431 (8 OCTOBER 1972)
Lake Name
Carlos
Le Homme Dieu
Darling
Mi n new ask a
Nest
Green
Wagon ga
Winona
Trace
o Big Stone
rsi
Serial
Number
17
18
16
19
20
21
22
101
102
104
Pixel
Count
2,200
1,561
923
7,028
872
4,986
1,501
425
102
10,407
LANDSAT-1 Lake
Area (ha)
1,050.9
745.7
440.9
3,357.3
416.6
2,381.8
717.0
203.0
89.8
4,971.4
Map Lake
Area (ha)
1,019.8
705.8
386.1
2,877.4
382.4
2,187.8
654.4
73.3
5,103.3
Map Area:LANDSAT-l
Area Ratio
0.970
0.947
0.876
0.857
0.918
0.919
0.913,
a
1.027
JThe LANDSAT-1 image includes pixels from Lake Agnes and Lake Henry.
-------
TABLE D7-5.
MSS DESCRIPTIVE STATISTICS FOR 10 MINNESOTA LAKES
EXTRACTED FROM FRAME 1077-16431 (8 OCTOBER 1972)
Lake Name
Carlos
Le Homme Dieu
Darling
Minnewaska
Nest
Green
Wagonga
Trace
Winona
Big Stone
Serial
Number
17
18
16
19
20
21
22
102
101
104
Pixel
Count
2,200
1,561
923
7,028
872
4,986
1,501
188
425
10,407
LANDSAT-1 MSS Bands
Green
33.19a
2.86b
33.27
2.56
31.72
1.91
37.48
1.68
31.94
1.95
33.69
2.27
40.97
2.47
36.20
4.62
35.02
2.40
38.76
1.93
Red
17.25
3.36
17.54
2.95
16.61
2.76
20.11
2.07
17.73
2.26
17.16
2.12
24.96
2.11
21.38
5.21
19.92
3.09
22.56
3.05
IR1
10.58
4.56
10.48
4.16
10.14
4.29
11.14
3.05
11.53
4.87
9.35
2.81
24.91
3.20
15.68
7.02
14.76
6.02
17.58
3.78
IR2
4.74
4.50
4.64
4.173
3.46
5.79
3.87
3.20
9.28
4.96
9.49
8.05
7.57
6.91
6.11
4.50
^Mean DN value for the lake.
bStandard deviation of the lake DN values.
203
-------
APPENDIX D8. 28 MAY 1973 (1309-16325)
Section D8.1. Regression Models, and Correlation Coefficients
(Table D8-1)
Thirteen Minnesota lakes were extracted from the frame. Lakes Calhoun
and Maple were not used in the development of the regression model.
Clearwater (23) Carrigan (27) Spring (33)
Maple (24) Silver (28) Madison (35)
Cokato (25) Forest (30) Sakatah (36)
Buffalo (26) White Bear (31) Calhoun (103)
St. Croix (32)
TABLE D8-1. CORRELATIONS BETWEEN MSS DATA (COLORS AND COLOR
RATIOS) AND PCI VALUES FOR 11 MINNESOTA LAKES
EXTRACTED FROM FRAME 1309-16325 (28 MAY 1973)
PCI
GRN
RED
IR1
IR2
GRNRED
6RNIR1
GRNIR2
REDIR1
REDIR2
IR1IR2
-0.068
-0.040
-0.014
-0.259
0.061
0.017
0.316
-0.036
0.434
0.702
The best model for the prediction of trophic state is:
PCI = -16.537 + 9.844 IR1IR2
This model explains about 49 percent of the variation about the mean
(Table D8-2) and is not adequate. The observed and predicted PCI values
are in Table D8-3.
204
-------
TABLE D8-2.
ANALYSIS OF VARIANCE TABLE OF THE PCI REGRESSION
MODEL FOR 11 MINNESOTA LAKES EXTRACTED FROM FRAME
1309-16325 (28 MAY 1973)
Analysis of Variance
Source
df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
10
1
9
41.326
20.385
20.941
4.133
20.385
2.327
8.761
R2 = 0.4933 x 100 = 49.33% s.e. of estimate = 1.525
TABLE D8-3. PCI RESIDUALS OF 11 MINNESOTA LAKE EXTRACTED
FROM FRAME 1309-16325 (28 MAY 1973)
Lake Name
Clearwater
Cokato
Buffalo
Carrigan
Silver
Forest
White Bear
St. Croix
Spring
Madison
Sakatah
Serial
Number
23
25
26
27
28
30
31
32
33
35
36
PCI
0.01
1.61
2.31
4.40
4.78
-1.22
-1.41
-0.17
2.33
1.36
1.38
pel
2.88
2.07
2.29
1.78
3.89
-0.19
-0.04
-0.82
1.65
1.51
0.38
PCI -PCI
-2.87
-0.46
0.02
2.62
0.89
-1.03
-1.34
0.65
0.68
-0.15
1.00
Section D8.2. Three-dimensional Color Ratio Model
The MSS color ratio model is displayed in Figure D8-1. There appears
to be little agreement between lake position in the model and trophic
state as defined by the PCI value.
Section D8.3. Concatenation of Extracted Lakes
The concatenation of 13 lakes is displayed as Figure D8-2. Sakatah
Lake includes both Upper Sakatah Lake and Lower Sakatah Lake.
205
-------
Section D8.4. MSS-Lake Surface Area Relationships
The areal aspects of the lakes are in Table D8-5.
Section D8.5. Lake MSS Descriptive Statistics
The MSS statistics are in Table D8-4.
TABLE D8-4. DESCRIPTIVE STATISTICS FOR 13 MINNESOTA LAKES
EXTRACTED FROM FRAME 1309-16325 (28 MAY 1973)
Lake Name
Clearwater
Maple
Cokato
Buffalo
Carrigan
Silver
Forest
White Bear
St. Croix
Spring
Madison
Sakatah
Calhoun
Serial
Number
23
24
25
26
27
28
30
31
32
33
35
36
103
Pixel
Count
2,689
584
442
1,327
401
368
1,871
2,093
2,133
4,856
1,177
977
339
LANDSAT-1 MSS Bands
Green
45.89*
3.24°
49.09
3.04
44.73
2.47
51 . 01
2.06
48.12
2.37
51.79
1.71
52.71
6.32
53.58
2.31
53.50
2.11
61.44
2.89
50.78
2.10
53.09
2.16
51.72
1.95
Red
26.74
2.94
28.11
2.48
26.60
2.08
31.60
2.17
28.29
2.12
31.92
1.58
33.06
2.59
32.92
2.80
37.24
2.62
47.49
3.61
31.15
2.19
35.79
2.98
30.17
2.21
IR1
19.09
3.92
21.15
4.40
20.32
3.51
22.04
2.82
20.86
3.49
28.36
3.04
26.18
3.04
26.85
3.47
30.31
2.76
37.79
2.91
22.67
3.33
27.31
3.27
23.93
2.89
IR2
9.68
4.18
11.19
4.91
10.75
4.05
11.52
3.24
11.21
4.26
13.67
3.80
15.76
3.31
16.02
3.15
18.98
2.77
20.46
3.01
12.36
3.89
15.90
3.78
13.81
3.32
bMean DN value for the lake.
Standard deviation of the lake DN values.
206
-------
2.33
ro
o
-0.17
St. Croix
4.79
v Silver
1,38
\Sakatah
-I .41
White Be
-1.22
Forest
I .61
CokatoQ
Calhoun ,
2.31
. Buffalo
I .36
•^Madison
t . "i u f~*
Car riganuy
0.01 /O
C|earwater(2
->JJlaple
G R N : I R1
Figure D8-1.
Three-dimensional MSS color ratio model of 13 Minnesota lakes extracted from
Frame 1309-16325 (28 May 1973).
-------
Clear rt a M
Cokalo
For e »'
"" 27
C o r r i g a n
Saha tah
a i so n
28
Silver
MINNESOTA
Figure D8-2.
IR2 concatenation of 13 Minnesota lakes extracted from
Frame 1309-16325 (28 May 1973).
208
-------
TABLE D8-5. AREAL ASPECTS OF 13 MINNESOTA LAKES EXTRACTED FROM FRAME 1309-16325 (28 MAY 1973)
o
|JD
Lake Name
Clearwater
Maple
Cokato
Buffalo
Carrigan
S i 1 ver
Forest
White Bear
St. Croix
Spring
Madison
Sakatah
Calhoun
Serial
Number
23
241
25
26
27
28
30
31
32
33
35
36
103
Pixel
Count
2,689
584
442
1,327
401
368
1,871
2,093
2,133
4,856
1,177
977
339
LANDSAT-1 Lake
Area (ha)
1,284.5
279.0
211.1
633.0
191.6
175.8
893.8
999.8
1,018.9
2,319.7
562.3
466.7
161.9
Map Lake
Area (ha)
1,287.8
287.3
220.2
611.1
65.0
170.8
892.8
1,076.5
3,322.2
2,391.8
541.1
497.0
169.6
Map Area: LANDSAT-1
Area Ratio
1.003
1.030
1.043
0.9653
a
0.972
0.999
1.077b
U
1.031
0.962
1.065
1.048
?A portion of the lake was outside the sensor field of view.
The entire lake image was not extracted from the CCT's.
-------
APPENDIX D9. 3 July 1973 (1345-16322)
Section D9.1. Regression Models, and Correlation Coefficients
(Table D9-1)
Fourteen Minnesota lakes were extracted from the frame. Lakes Calhoun
and Maple were not used in developing the regression model.
Clearwater (23) Silver (28) St. Croix (32)
Maple (241) Minnetonka (29) Spring (33)
Cokato (25) Forest (30) Madison (35)
Buffalo (26) White Bear (31) Sakatah (36)
Carrigan (27) Calhoun (103)
TABLE D9-1. CORRELATIONS BETWEEN MSS DATA (COLOR RATIOS)
AND PCI VALUES FOR 12 MINNESOTA LAKES EXTRACTED
FROM FRAME 1345-16322 (3 JULY 1973)
PCI
GRNRED 0.239
GRNIR1 -0.788
GRNIR2 -0.641
REDIR1 -0.837
REDIR2 -0.642
IR1IR2 0.172
The best model for the prediction of trophic state is:
PCI = 10.544 - 7.240 REDIR1
This model explains about 70 percent of the variation about the mean
(Table D9-2), but is of little practical value. The observed and pre-
dicted PCI values are in Table D9-3.
210
-------
TABLE D9-2.
ANALYSIS OF VARIANCE TABLE OF THE PCI REGRESSION
MODEL FOR 12 MINNESOTA LAKES EXTRACTED FROM FRAME
1345-16322 (3 JULY 1973)
Analysis of Variance
Source
df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
11
1
10
41.744
29.260
12.484
3.795
29.260
1.248
23.438
= 0.7009 x 100 = 70.09% s.e. of estimate = 1.117
TABLE D9-3. PCI RESIDUALS OF 12 MINNESOTA LAKES EXTRACTED FROM
FRAME 1345-16322 (3 JULY 1973)
Lake Name
Clearwater
Cokato
Buffalo
Carrigan
Silver
Minnetonka
Forest
White Bear
St. Croix
Spring
Madison
Sakatah
Serial
Number
23
25
26
27
28
29
30
31
32
33
35
36
PCI
0.01
1.61
2.31
4.40
4.79
0.73
-1.22
-1.40
-0.17
2.33
1.36
1.38
PCI
-0.08
0.33
3.05
3.13
4.90
-0.31
-0.54
0.75
0.76
1.25
0.98
1.92
PCl-PCl
0.09
1.28
-0.74
1.27
-0.11
1.04
-0.68
-2.15
-0.93
1.08
0.38
-0.54
Section D9.2. Three-dimensional Color Ratio Model
The MSS color ratio model is displayed in Figure D9-1. Silver Lake's
IR1 DN level exceeds its RED DN level; this isolates it from the other
lakes.
Section D9.3. Concatenation of Extracted Lakes
The concatenation of the 14 lakes is in Figure D9-2.
211
-------
Section D9.4. MSS-Lake Surface Area Relationships
The areal aspects of the 14 lakes are in Table D9-5.
Section D9.5. Lake MSS Descriptive Statistics
The MSS statistics are in Table D9-4.
TABLE D9-4. MSS DESCRIPTIVE STATISTICS FOR 14 MINNESOTA LAKES
EXTRACTED FROM FRAME 1345-16322 (3 JULY 1973)
Lake Name
Clearwater
Maple
Cokato
Buffalo
Carrigan
Silver
Minnetonka
Forest
White Bear
St. Croix
Spring
Madison
Sakatah
Calhoun
jMean DN value
Serial
Number
23
241
25
26
27
28
29 1
30
31
32
33
35
36
103
for the lake.
Pixel
Count
2,485
505
437
1,285
106
346
1,239
1,818
2,094
2,145
4,945
1,151
986
342
LANDSAT-1 MSS Bands
Green
47.29?
3.25°
43.43
2.49
41.50
2.45
47.77
2.01
47.41
2.04
50.44
1.68
47.05
3.79
42.24
2.27
40.94
3.05
38.78
2.26
43.93
2.51
46.84
2.23
44.46
2.25
50.45
2.22
Red
26.59
2.49
24.15
2.21
22.54
2.27
26.01
1.63
25.16
1.37
27.61
1.25
25.44
3.26
24.61
2.24
22.78
3.27
22.50
2.77
28.36
2.27
27.10
1.70
26.43
1.87
24.91
2.57
IR1
18.11
9.85
16.94
7.55
15.98
3.67
25.12
4.68
24.56
4.44
35.42
3.30
16.96
9.07
16.07
3.99
16.84
8.33
16.66
7.24
22.09
1.68
20.51
8.51
22.18
4.13
15.65
3.75
IR2
7.21
5.19
7.86
5.43
6.88
4.42
9.04
3.72
13.35
3.92
13.43
3.94
6.89
4.66
7.82
4.22
7.94
4.41
6.59
4.26
8.03
4.55
7.79
4.63
9.22
4.40
6.29
4.13
values.
212
-------
ro
oo
G R N : I R 1
Figure D9-1. Three-dimensional MSS color ratio model of 14 Minnesota lakes extracted from
Frame 1345-16322 (3 July 1973).
-------
29
Mmnetonka
35
ad i »o n
25
Cokalo
30
F o r c 11
27
o rngan
23
C I eo r w ot«
24 I
M o p t *
26
I u I (o|o
31
While Bear
32
SI Cro,.
103
C a I hou r
36
So k a IO h
IHIECT
MINNESOTA
Figure D9-2.
IR2 concatenation of 14 Minnesota lakes extracted from
Frame 1345-16322 (3 July 1973).
214
-------
<_n
TABLE D9-5. AREAL ASPECTS OF 14 MINNESOTA LAKES EXTRACTED FROM FRAME 1345-16322 (3 JULY 1973)
Lake Name
Clearwater
Maple
Cokato
Buffalo
Carrigan
Silver
Minnetonka
Forest
White Bear
C f r rn i Y
O L . wl U 1 A
Spring
Madison
Sakatah
Calhoun
Serial
Number
23
241
25
26
27
28
29
30
31
OL.
33
35
36
103
Pixel
Count
2,485
505
437
1,285
106
346
11,239
1,818
2,094
2 1 45
4^945
1,151
986
342
LANDSAT-1 Lake
Area (ha)
1,187.1
241.2
208.8
613.8
50.6
165.3
5,368.9
868.5
1,000.3
1 024 7
1 j \s t. T • /
2,362.2
549.8
471.0
163.4
Map Lake
Area (ha)
1,287.8
287.3
220.2
611.1
65.6
170.8
5,855.6
892.8
1,076.5
3 322 2
\J ) *J L, L_ • ^
2,391.8
541.1
497.0
169.6
Map Area:LANDSAT-l
Area Ratio
1.085
1.191
1.055
0.996
1.296
1.033
1.091
1.028
1.076a
a
1,013
0.984
1.055
1.038
aThe entire lake image was not extracted from the CCT's.
-------
APPENDIX D10. 4 July 1973 (1346-16381)
Section D10.1.
Regression Models, and Correlation Coefficients
(Table D10-1)
Eight Minnesota lakes were extracted from the frame;
was not used in developing the regression model.
Cottonwood Lake
Nest (20)
Green (21)
Wagonga (22)
Cokato (25)
TABLE D10-1.
Buffalo (26)
Carrigan (27)
Silver (28)
Cottonwood (111)
CORRELATIONS BETWEEN MSS DATA (COLORS AND COLOR
RATIOS) AND PCI VALUES FOR 7 MINNESOTA LAKES
EXTRACTED FROM FRAME 1346-16381 (4 JULY 1973)
PCI
GRN
RED
IR1
IR2
GRNRED
GRNIR1
GRNIR2
REDIR1
REDIR2
IR1IR2
0.874
0.829
0.903
0.829
-0.595
-0.956
-0.800
-0.961
-0.839
0.573
The best model for the prediction of lake trophic state is:
PCI = 11.715 - 8.277 REDIR1
This model accounts for about 92 percent of the variation about the
mean (Table D10-2). The observed and predicted PCI values are in
Table D10-3.
216
-------
TABLE D10-2. ANALYSIS OF VARIANCE TABLE OF THE PCI REGRESSION
MODEL FOR 7 MINNESOTA LAKES EXTRACTED FROM FRAME
1346-16381 (4 JULY 1973)
Analysis of Variance
Source
df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
6
1
5
29.233
26.994
2.238
4.872
26.994
0.448
60.255
R2 = 0.9234 x 100 = 92.34% s.e. of estimate = 0.669
TABLE D10-3. PCI RESIDUALS OF 7 MINNESOTA LAKES EXTRACTED
FROM FRAME 1346-16381 (4 JULY 1973)
Lake Name
Nest
Green
Wagonga
Cokato
Buffalo
Carrigan
Silver
Serial
Number
20
21
22
25
26
27
28
PCI
0.77
-1.10
4.40
1.61
2.31
4.40
4.79
XV.
PCI
0.33
-0.34
4.04
0.86
3.07
4.71
4.49
PCI -PCI
0.44
-0.76
0.36
0.75
-0.76
-0.31
0.30
Section DIP.2. Three-dimensional Color Ratio Model
The MSS color ratio model is displayed in Figure D10-1. Lakes Wagonga,
Carrigan, and Silver have IR1 values which exceed their RED DN values.
Section 10.3. Concatentation of Extracted Lakes
The concatenation of eight lakes is in Figure D10-2.
Section 10.4. MSS-Lake Surface Area Relationships
The areal aspects of the eight lakes are in Table D10-4.
Section DIP.5. Lake MSS Descriptive Statistics
The MSS statistics are in Table D10-5.
217
-------
ro
co
1.61
2.31
/— \Buffalo
4.79
Silver
Qp Cottonwood
f"\ I fllll
^c ' T
4r4'5an
4.40
Wagonga
©
Cokato
0.77
)Nest (TT)G
-I.10
Green
G R N : I R 1
Figure D10-1. Three-dimensional MSS color ratio model of 8 Minnesota lakes extracted from
Frame 1346-16381 (4 July 1973).
-------
2 7
Cor rIg o n
S.lv
2 5
C o I o I o
1 1 1
Cottonwood
W o g o n go
21
Green
ft
Bv.Holo
MINN
Figure D10-2.
IR2 concatenation of 8 Minnesota lakes extracted from
Frame 1346-16381 (4 July 1973).
219
-------
TABLE D10-4. AREAL ASPECTS OF 8 MINNESOTA LAKES EXTRACTED FROM FRAME 1346-16381 (4 JULY 1973)
ro
IX)
o
Lake Name
Nest
Green
Wagonga
Cokato
Buffalo
Carrigan
Silver
Cottonwood
Serial
Number
20
21
22
25
26
27
28
111
Pixel
Count
724
4,701
1,275
434
1,300
98
346
293
LANDSAT-1 Lake
Area (ha)
345.9
2,245.7
609.1
207.3
721.0
46.8
165.3
140.0
Map Lake
Area (ha)
382.4
2,187.8
654.4
220.2
611.1
75.6
170.8
149.7
Map Area: LANDSAT-1
Area Ratio
1.106
0.974
1.074
1.062
0.984
1.402
1.033
1.069
-------
TABLE D10-5.
MSS DESCRIPTIVE STATISTICS FOR 8 MINNESOTA LAKES
EXTRACTED FROM FRAME 1346-16381 (4 JULY 1973)
Lake Name
Nest
Green
Wagonga
Cokato
Buffalo
Carrigan
Silver
Cottonwood
Serial
Number
20
21
22
25
26
27
28
111
Pixel
Count
724
4,701
1,275
434
1,300
98
346
293
LANDSAT-1 MSS Bands
Green Red IR1
45.37*
2.86°
47.34
3.48
58.42
2.79
52.32
3.83
59.55
3.90
59.77
3.37
60.43
3.67
52.35
1.46
27.24
2.38
27.54
2.93
35.47
3.09
35.20
4.22
39.73
4.23
39.97
3.42
40.78
4.02
32.04
1.38
19.79
3.61
18.91
2.71
38.25
3.04
26.85
3.72
38.07
4.13
47.26
3.73
46.73
3.33
30.10
3.50
IR2
10.34
4.12
10.06
3.08
15.45
3.39
16.04
4.02
18.97
3.52
24.08
3.28
22.61
3.39
12.13
3.91
.Mean DN value for the lake.
Standard deviation of the lake DN values.
221
-------
APPENDIX Dll. 19 August 1972 (1027-15233)
Section D11.1. Regression Models, and Correlation Coefficients
(Table Dll-1)
Seven New York lakes were extracted from the frame; all were incor-
porated into the regression model.
Conesus (91)
Canandaigua (92)
Keuka (93)
Seneca (94)
Cayuga (95)
Owasco (96)
Cross (97)
TABLE Dll-1. CORRELATIONS BETWEEN GROUND TRUTH AND MSS DATA
(COLORS AND COLOR RATIOS) FOR 7 NEW YORK LAKES
EXTRACTED FROM FRAME 1027-15233 (19 AUGUST 1972)
PCI
CHLA
LNCHLA SECCHI LNSECCHI
GRN
RED
IR1
IR2
GRNRED
GRNIR1
GRNIR2
REDIR1
REDIR2
IR1IR2
0.245
0.772
0.752
0.890
-0.906
-0.596
-0.740
-0.286
-0.645
-0.860
0.601
0.833
0.632
0.599
-0.717
-0.331
-0.337
-0.032
-0.207
-0.349
0.463
0.664
0.550
0.510
-0.580
-0.307
-0.284
-0.084
-0.186
-0.268
-0.792
-0.507
0.063
0.116
0.072
-0.419
-0 . 389
-0.621
-0.490
-0.239
-0.804
-0.572
-0.013
0.023
0.158
-0.356
-0.315
-0.583
-0.423
-0.152
The best model for the prediction of the lake trophic state index is:
PCI - -4.981 - 8.805 GRNIR1 + 19.301 REDIR1
This model explains about 83 percent of the variation about the mean
(Table Dll-2). The observed and predicted PCI values are in Table
Dll-3.
222
-------
TABLE Dll-2. ANALYSIS OF VARIANCE TABLE OF THE PCI REGRESSION MODEL
FOR 7 NEW YORK LAKES EXTRACTED FROM FRAME 1027-15233
(19 AUGUST 1972).
Source Analysis of Variance
df Sum of Squares Mean Square Calculated F
Total (corrected)
Regression
Residual
6
2
4
12.741
10.552
2.189
2.124
5.276
0.547
9.645
R2 = 0.8282 x 100 = 82.82% s.e. of estimate = 0.740
TABLE Dll-3. PCI RESIDUALS OF 7 NEW YORK LAKES EXTRACTED FROM
FRAME 1027-15233 (19 AUGUST 1972).
Lake Name Serial
Cones us
Canandai
Keuka
Seneca
Cayuga
Owasco
Cross
Section Dll. 2.
Number
91
gua 92
93
94
95
96
97
Three-dimensional
PCI
-1.41
-3.63
-2.14
-2.89
-2.74
-2.47
0.86
Color Ratio
PCI
-1.82
-2.67
-2.82
-2.58
-3.28
-1.98
0.73
Model
PCI -PCI
0.41
-0.96
0.68
-0.31
0.52
-0.49
0.13
The MSS color ratio is displayed in Figure Dll-1. Lake Canandaigua
appears to be misplaced if its PCI value is an accurate assessment
of its trophic state.
Section D11.3. Concatenation of Extracted Lakes
The concatenation of seven lakes is found in Figure Dll-2.
Section Oil.4. MSS-Lake Surface Area Relationships
The area! aspects of the seven lakes are in Table Dll-4.
Section D11.5. Lake MSS Descriptive Statistics
The MSS statistics are in Table Dll-5.
223
-------
ro
ro
0.86
-1.41
,Conesus
-3.63
Canandaigua
-2.47
v 0 wasco
-2.89
Seneca (M)
-2.14
Keuka
-2.74
C ay uga (95)
G R N : I R 1
Figure Dll-1.
Three-dimensional MSS color ratio model of 7 New York lakes extracted from Frame
1027-15233 (19 August 1972).
-------
nondo •9 « o
i
Figure Dll-2.
IR2 concatenation of 7 New York lakes extracted from
Frame 1027-15233 (19 August 1972).
225
-------
TABLE D11-4. AREAL ASPECTS OF 7 NEW YORK LAKES EXTRACTED FROM FRAME 1027-15233 (19 AUGUST 1972).
Lake Name
Conesus
Canandaigua
Keuka
Seneca
Cayuga
Owasco
Cross
Serial
Number
91
92
93
94
95
96
97
Pixel
Count
2,764
9,009
9.896
37,782
37,339
5,835
1,712
LANDSAT-1
Lake Area (ha)
1,319.9
4,302.2
4,725.8
17,583.9
17,831.1
2,786.5
817.6
Map Lake
Area (ha)
1,347.3
4,219.8
4,739.9
17,252.5
17,319.9
2,745.5
844.4
Map Area: LANDSAT-1
Area Ratio
1.021
0.998
1.003
0.981
0.971
0.985
1.033
ro
ro
en
-------
TABLE Dll-5.
MSS DESCRIPTIVE STATISTICS FOR 7 NEW YORK LAKES EXTRACTED
FROM FRAME 1027-15233 (19 AUGUST 1972).
Lake Name
Conesus
Canandaigua
Keuka
Seneca
Cayuga
Owasco
Cross
Serial
Number
91
92
93
94
95
96
97
Pixel
Count
2,764
9,009
9,896
37,782
37 ,339
5,835
1,712
LANDSAT-1 MSS Bands
Green
37.38*
2.12b
40.17
3.14
38.49
2.32
42.53
2.68
41.11
1.92
40.79
4.00
43.25
3.62
Red
19.67
2.21
20.20
2.68
19.09
2.19
21.02
1.99
19.86
3.01
20.78
4.05
25.34
3.16
IR1
16.47
3.49
16.32
2.04
14.23
3.37
13.49
3.61
13.25
3.41
14.38
3.72
19.26
4.10
IR2
7.83
4.14
6.05
3.83
5.99
4.09
5.04
3.45
5.11
3.42
5.94
3.81
9.11
4.68
DN value for the lake.
Standard deviation of the lake DN values
227
-------
APPENDIX D12. 11 October 1972 (1080-15180)
Section D12.1. Regression Models and Correlation Coefficients
Five lakes were extracted from the frame. No regression models were
constructed due to the small number of observations. Correlation
coefficients between ground truth and MSS data were not determined for
the same reason.
Section D12.2. Three-dimensional Color Ratio Model
The MSS color ratio model is displayed in Figure D12-1. The large
shift in position of Lake Cayuga (95) - compare with Figure 58 - may
be a consequence of the thin cloud deck over it.
Section D12.3. Concatenation of Extracted Lakes
The extracted lakes are in Figure D12-2. Lakes Cross and Cayuga are
only partially in the sensor field of view; this accounts for the
linear "shore lines".
Section D12.4. MSS-Lake Surface Area Relationships
The areal aspects of the lakes are in Table D12-1.
Section D12.5. Lake MSS Descriptive Statistics
The MSS statistics are in Table D12-2.
228
-------
ro
ro
-2.47
G R N :| R 1
Figure D12-1. Three-dimensional MSS color ratio model of five New York lakes extracted from Frame
1080-15180 (11 October 1972).
-------
O W O t
107
C onodorogo
97
C r o i *
1080-15180-L-5 IP£ 110CT72 UPPER NEU YORK
Figure D12-2.
IR2 concatenation of five New York lakes extracted from
Frame 1080-15180 (11 October 1972).
230
-------
TABLE D12-1. AREAL ASPECTS OF 5 NEW YORK LAKES EXTRACTED FROM FRAME 1080-15180 (11 OCTOBER 1972).
Lake Name
Cayuga
Owasco
Cross
One i da
Canadarago
Serial
Number
95
96
97
106
107
Pixel
Count
8,799
6,018
1,072
44,934
1,644
LANDS AT -1
Lake Area (ha)
4,206.0
2,876.6
512.4
21,478.5
785.8
Map Lake
Area (ha)
17,319.9
2,745.5
844.4
20,720.6
Map Area: LANDSAT-1
Area Ratio
a
0.955
a
0.963
Only a fraction of the lake surface is in the sensor field of view.
CO
-------
TABLE D12-2.
MSS DESCRIPTIVE STATISTICS FOR 5 NEW YORK LAKES EXTRACTED
FROM FRAME 1080-15180 (11 OCTOBER 1972).
Lake Name
Cayuga
Owasco
Cross
Oneida
Canadarago
Serial
Number
95
96
97
106
107
Pixel
Count
8,799
6,018
1,072
44,934
1,664
LANDSAT-1 MSS Bands
Green
36.31-*
2.72b
38.66
2.94
38.68
1.28
34.89
2.11
35.53
2.18
Red
18.18
2.89
19.46
3.63
22.63
3.01
18.07
2.15
19.89
2.62
IR1
12.49
4.88
11.40
3.75
13.58
3.87
10.53
2.77
11.75
4.20
IR2
6.23
5.63
4.53
4.00
4.93
4.68
4.17
2.93
4.63
Mean DN value for the lake.
Standard deviation of the lake DN values.
232
-------
APPENDIX E
N x N SQUARED EUCLIDIAN DISTANCE MATRIX
The dendrogram of 100 NES-sampled lakes (Figure 15) was created using
the output of the McKeon hierarchical cluster analysis program. The
clustering procedure was carried out using the matrix in this appendix.
Only the lower triangular form of the matrix is reproduced here.
MCKEON CLUSTER ANALYSIS VERSION I.I
CONTROL INPUT: 100 3 6 3 2 1 2
:<3X,6F10.5)
FIRST ROW OF DATA FOR VERIFICATION-
2.678E 00 -5.413E-01 5.497E 00 -2.976E 00 -2_.806E-01_ 7.655E-01
SQUARED DISTANCES BETWEEN POINTS BEFORE CLUSTERING
1
2
3
4
5
6
7
8
9
10
11
12
13
14
IS
16
17
18
19
20
21
2?
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
1
0
.633
.511
1.052
1.191
2.536
2.258
2.244
1.564
1.319
4.698
5.546
.943
1.646
4.553
2.03?
3.929
2.501
4.325
.936
2.998
22.133
.657
22.071
6.173
5.819
21.351
23.864
1.530
2.564
4.547
1.42?
8.760
5.909
2.281
4.586
12.202
36.761
1.373
1. 169
8.347
7.268
12.322
29.181
2.714
1.149
4.370
3.220
6.350
6.285
5.998
3.211
1 1.689
3.046
2
0
.728
1.490
.571
1.271
1.SS1
.760
2.666
2.450
6.044
7.828
1.131
1.209
6.822
1.002
1.900
1.469
2.385
1.898
1.033
25.375
.597
25.930
6.436
8.049
25.121
29.064
2.119
1.555
3.223
2.689
10.724
7.541
4.371
5.083
15.716
42.864
2.334
2.111
10.062
9.988
10.826
26.758
4.360
1.078
6.701
5.337
9.497
3.503
3.099
3.118
15.151
4.377
3
0
.505
.542
1.730
1.658
1.756
1.439
2.050
4.580
6.684
.896
2.736
S.087
1.856
2.991
2.072
3.705
1.477
2.487
23.284
.489
23.830
7.887
7.981
23.003
26.324
2.532
2.020
5.379
1.360
8.534
7.815
3.025
6.512
12.798
38.964
1.313
1.211
8.767
8.649
11.275
27.907
3.246
.938
4.492
2.837
7.102
5.557
5.836
3.972
12.904
3.487
4
0
1.511
3.307
2.961
2. 522
.431
2.568
5.190
6.858
1.774
4.672
5.238
3.473
5.214
3.294
4.818
1.840
3.402
19.575
1.350
19.299
7.886
7.092
19.342
23.642
3.000
3.166
8.133
1.635
6.602
7.722
2.886
6.034
10.813
34.615
1.078
1.258
9.644
8.696
13.706
31.161
2.756
2.255
4.095
2.169
5.545
7.835
7.215
3.857
10.344
2.318
5
0
.395
.695
.951
3.280
2.589
4.318
7.412
.785
2.396
5.425
1.110
1.409
.851
3.191
3.097
1.436
29.635
1.129
30.077
10.346
11.475
29.424
33.403
3.684
.691
3.440
2.430
12.953
11.043
5.592
8.679
17.877
47.677
2.183
2.145
8.384
10.007
8.212
22.466
5.223
.340
7.917
5.815
11.356
2.812
3.665
5.566
18.115
6.202
6
a
.679
1.244
5.833
3.605
4.884
8.634
1.617
2.571
6.758
1.441
.696
.952
3.481
5.192
1.375
35.780
2.342
36.507
12.829
15.212
35.476
39.725
5.317
.630
2.394
3.739
16.994
13.889"
8.361
11.067
22.567
55.855
3.748
3.857
8.271
11.161
6.323
18.838
7.487
.692
10.875
8.777
15.559
1.290
3.065
7.443
23.223
9.451
7
0
.962
5,.290
2.155
2.840
5_,540
.566
3.070
6.605
2.634
1.552
2.373
6.090
5.376
1.497
36.401
2.727
36.287
12.407
14.650
35,836
40.050
6.562
1.885
4.110
3.467
15.817
14.202
8.564
10.305
23.367
55.514
2.074
2.~025
8.426
9.036
4.519
16.490
4.854
.972
10.778
8.212
14.697
3.058
3.122
6.252
22.B82
8.760
.8
0
4.401
3.515
6.188
8.561
1.314
2.707
9.630
1.968
1.361
2.379
3.777
4.397
.204
31.353
1.980
31.833
8.642
11.985
31.508
37.009
5.317
2.275
4.569
3.829
13.246
11.679
8.050
7. 167
21.542
51.164
2.811
2.491
12.584
12.184
8.003
22.066
5.083
2.012
10.326
8.045
13.484
2.949
1.344
3.784
20.228
6.617
9
0
3.528
7.145
7.854
3.177
5.927
6.036
5.008
7.975
5.216
6.143
1.444
5.512
14.676
2.011
14.242
6.5S6
4.604
14.417
18.172
2.742
5.310
10.609
2.453
4.447
5.839
1.738
4.890
7.558
27.557
1.818
1.853
11.652
9.345
18.326
37.900
2.640
3.986
2.796
1.242
3.083
11.291
9.885
3.659
6.717
.947
10
0
1.847
1.642
1.278
3.232
4.071
5.363
5.845
5.128
8.991
4.005
4.293
28.494
3.337
26.961
9.523
9.423
26.603
29.290
4.821
4.457
6.205
.522
11.355
9.154
5.215
6.644
16.692
43.687
.746
1.208
5.286
3.003
8.577
22.704
1.726
3.280
6.572
5.316
9.105
7.933
7.664
4.703
15.684
6.131
11
0
1.421
2.697
7.943
3.375
8.718
8.189
6.931
14.415
9.164
7.498
39.817
7.454
36.974
19.213
17.974
37.647
40.641
10.684
5.688
9.321
2.256
18.630
18.537
10.509
15.121
24.474
55.890
2.241
3.100
4.148
4.073
5.147
15.198
4.825
3.551
12.338
9.341
15.342
9.643
10.679
11.341
24.023
12.033
12
0
4.419
8.772
5.346
12.168
12.181
11.130
18.137
9.789
9.894
36.583
9.335
32.902
16.559
15.137
33.579
35.882
11.343
9.832
12.313
3.636
16.683
15.790
10.454
12.203
23.099
50.561
2.732
3.631
5.601
1,778
8.843
20.440
3.247
6.517
11.622
9.592
13.567
14.419
13.600
9.857
21.256
11.459
13
0
2.622
5.156
2.370
2.790
2.526
6.215
3.156
2.281
30.258
1.713
29.806
9.711
10.320
29.636
32.866
4.669
2.323
4.862 "
1.578
12.429
10.899
5.554
7.782
18.586
46.610
1.016
\657
8.733
7.980
7.279
21.185
2.945
.798
7.966
5.610
10.444
5.125
4.285
4.686
17.644
5.702
14
0
8.601
1.741
2.568
3.021
3.638
2.907
3.689
29.357
1.802
30.524
6.218
8.627
28.481
30.494
2.445
3.261
1.609
4.316
14.238
6.965
5.548
5.311
18.445
46.503
4.837
4.315
10.557
10.196
11.877
37.697
6.309
3.191
8.348
8.041
13.016
4.043
4.345
4.197
18.125
7.119
15
0
8.134
11.240
5.661
12.471
7.214
11.161
30.971
7.334
28.033
19.590
14.305
38.754
39.963
7.009
4.990
8.987
4.393
18.229
15.510
6.664
15.560
16.543
43.655
4.710
5.728
3.121
6.141
13.910
27.000
8.351
3.794
9.885
7.748
11.57r
11.675
15.287
14.366
17.746
10.479
233
-------
55
56
57
58
59
61
62
63
64
65
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
3.091
8.208
5.089
7.649
2.756
4.125
2.252
11.530
3. Ill
7.676
5.095
6.039
4.870
15.745
21.170
3.713
4.033
2.656
1.381
12.380
7.037
8.931
3.841
2.562
4.035
1.107
1.813
3.869
2.520
4.043
5.388
25.252
1.983
2.441
2.103
3.453
20.187
9.146
19.975
12.260
9.001
4.214
.810
2.549
1.541
16
0
1.386
.702
1.685
2.764
2.348
29.282
1.116
30.948
9.144
10.573
29.668
32.638
2.876
1.238
2.347
5.607
15.017
1.405
9.624
2.986
4.562
1.568
.378
4.405
2.495
13.762
1.170
4.485
6.972
7.416
2.448
11.669
16.064
2.687
3.334
1.495
1.159
14.671
8.526
5.796
4.706
2.126
3.785
.650
.985
4.269
2.177
4.103
6.541
23.701
3.329
2.220
3.063
1.759
15.770
7.447
14.725
8.077
6.702
3.641
1.123
4.488
1.785
17
0
1.996
3.091
6.084
1.226
38.305
2.739
40.444
12.010
16.371
38.631
42.921
6.504
2.009
2.586
6.024
18.404
2.293
10.169
3.651
7.192
2.264
.666
4.494
3.207
12.389
2.587
6.079
6.282
5.608
4.027
12.830
19.025
2.697
2.202
1.666
1.586
15.909
8.319
7.785
5.095
2.165
4.792
1.439
1.978
4.031
1.508
5.346
6.000
23.250
2.604
1.939
2.039
2.430
16.489
7.158
16.114
9.468
6.417
3.651
.595
2.806
1.575
18
0
2.458
4.223
2.785
31.824
2.244
32.318
12.858
13.277
31.841
35.564
3.783
.136
2.387
5.415
17.342
3.842
8.836
5.336
10.031
4.167
2.029
4.097
4.523
9.802
4.385
8.156
5.260
4.759
6.109
15.059
20.417
4.181
2.104
2.415
1.809
15.185
6.432
10.787
4.381
2.403
4.443
2.450
2.943
3.607
1.196
5.178
4.615
26.207
3.515
2.815
2.583
3.486
17.186
8.843
16.875
10.978
7.569
2.515
.326
2.228
2.666
19
0
4.032
3.320
24.940
2.349
27.824
8.385
11.031
25.792
30.272
2.964
3.078
4.408
8.483
13.694
.893
13.723
1.915
4.522
1.984
.345
6.933
4.140
17.274
1.368
3.188
9.647
7.555
1.850
8.457
14.144
1.051
3.078
1.813
2.547
20j,061
11.917
4.478
7.571
3.455
6.602
1.712
2.300
6.549
2.781
7.290
9.258
18.687
2.608
1.677
2.495
1.068
12.349
5.710
12.376
6.069
3.928
5.226
1.442
5.441
1.869"
20
0
5.268
15.775
.709
16.918
4.123
3.065
15.598
17.285
.799
4.849
7.046
3.466
6.044
.643
17.709
1.403
2.853
2.399
.654
9.477
5.514
22.154
1.134
1.845
13.461
10.051
.961
5.997
11.215
.633
4.668
2.772
4.038
24.045
15.846
2.368
10.541
5.312
8.855
2.847
3.430
9.200
4.709
9.488
12.849
15.566
3.352
2.368
3.710
1.184
10.675
5.460
11.162
4.296
3.243
7.711
2.927
8.461
2.922
21
0
32.522
2.518
33.366
8.674
13.054
32.649
38.617
5.661
2.615
4.234
4.669
14.160
2.250
16.927
3.691
5.616
2.850
1.145
8.761
6.898
21.519
2.385
3.448
12.542
12.318
2.5r
7.412
11.141
1.787
3.914
2.442
3.216
23.653
14.254
4.672
9.217
5.827
8.046
3.147
4.021
8.947
5.110
8.789
12.029
13.874
2.150
.631
1.501
1.194
10.606
3.325
12.874
6.208
3.470
7.971
3.192
6.535
1.319
22
0
21.452
1.193
14.082
8.076
.401
2.722
16.358
33.256
40.173
25.427
7.091
1.818
12.687
3.316
4.780
1.816
1.022
5.804
5.491
17.452
1.271
3.062
9.894
11.640
2.128
8.502
11.073
3.045
2.639
.983
1.637
19.561
10.698
5.134
6.353
3.401
4.691
1.758
2.013
6.079
3.156
5.608
8.761
20.152
4.561
1.301
2.985
.698
10.767
4.187
10.636
5.562
4.432
4.558
2.696
6.057
1.696
23
0
23.002
5.507
6.291
21.529
24.329
1.519
2.623
4.718
2.807
8.204
6.095
6.167
7.954
13.281
5.676
3.665
3.227
4.467
6.481
6.495
11.803
3.101
4.030
9.011
20.389
26.140
6.854
3.082
3.692
2.214
11.817
4.155
14.864
3.062
2.401
3.828
3.096
3.546
2.644
1.389
4.391
2.756
32.518
4.871
4.583
3.787
5.789
22.346
12.075
21.252
15.361
11.254
2.066
.815
1.355
3.829
24
0
16.184
8.540
.808
3.634
17.338
33.356
41.193
24.626
8.460
5.932
12.659
8.560
10.666
5.911
2.817
7.031
6.372
15.922
6.092
10.096
8.981
10.327
7.532
16.450
19.146
5.025
6.323
5.173
2.962
16.540
9.801
10.796
6.512
6.681
6.847
4.220
5.478
7.375
6.052
6.557
9.006
19.321
.642
2.209
1.071
5.406
20.451
8.825
24.083
14.256
10.109
8.410
2.354
3.925
2.031
25
0
2.530
14.044
16.850
4.767
13.830
13.446
9.588
5.716
8.278
22.389
10.384
14.168
10.057
5.803
14.515
12.752
25.254
9.695
10.935
16.387
14.479
9.627
13.643
17.160
4.980
9.098
8.792
7.884
28.537
17.975
11.808
13.793
12.569
14.380
8.962
10.925
14.721
10.592
14.663
16.646
11.366
.929
3.152
1.688
6.696
16.548
7.960
23.336
14.044
7.877
14.340
5.631
7.944
4.282
26
0
7.539
8.475
3.797
14.362
16.095
9.196
4.527
12.722
19.388
16.150
18.984
12.864
8.364
13.370
13.851
22.465
13.297
17.014
14.746
17.172
14.506
21.929
22.834
9.862
11.846
11.266
8.000
23.338
14.936
17.767
11.820
14.172
13.141
10.576
12.654
14.181
12.813
12.679
15.145
17.378
1.850
4.934
2.481
10.732
24.834
11.902
32.578
21.369
14.630
15.686
7.093
7.238
5.334
27
0
1.750
15.779
32.988
39.186
23.905
7.147
2.932
12.575
4.813
7.580
2.657
1.204
6.545"
5.087
16.825
2.979
5.620
8.491
10.054
3.946
1 1 . 344
15.942
2.713
3.425
2.171
2.103
18.612
10.460
7.338
6.219
4.337
6.048
2.015
2.987
6.611
3.889
6.623
8.650
18.195
1.400
.489
.645
1.743
14.391
4.671
16.040
9.533
5.424
6.192
2.193
3.694
.471
28
0
17.597
37.041
41.382
26.759
10.720
2.795
10.932
4.859
4.006
2.081
1.330
6.019
1.978
17.103
1.910
6.344
8.831
10.522
3.404
14.926
19.369
4.515
7.283
3.778
2.637
13.367
11.029
5.963
6.155
4.121
5.188
1.409
1.860
6.119
5.411
4.645
9.150
25.397
4.061
4.189
4.903
4.241
21.830
10.224
20.739
11.626
10.853
7.415
3.555
6.873
2.889
29
0
4.350
5.323
4.764
8.428
8.224
19.969
9.717
14.787
12.070
7.120
15.821
9.640
21.589
10.997
13.624
14.029
6.979
10.909
18.792
27.082
5.065
12.529
12.024
10.401
24.353
17.671
13.835
14.376
12.266
16.232
9.074
11.052
14.756
10.384
15.953
15.617
20.835
2.161
8.028
5.631
9.569
24.288
16.852
27.895
18.028
11.799
13.748
4.819
8.848
8.039
30
0
2.452
4.729
17.698
234
-------
34 10.198
35 5.749
36 8.631
37 18.389
38 46.942
39 5.304
40 4.463
41 12.821
42 15.241
43 12.363
44 28.144
45 8.450
46 1.240
47 9.308
48 7.719
49 12.593
50 2.781
51 3.547
52 6.354
53 18.800
54 6.820
55 .559
56 12.382
57 1.417
58 2.980
59 .986
60 .868
61 7.156
62 2.150
63 17.662
64 .537
65 3.106
66 9.113
67 7.920
68 1.236
69 10.296
70 17.845
71 2.148
72 4.801
73 2.256
74 3.404
75 17.814
76 12.631
77 4.109
78 7.569
79 2.792
80 6.487
81 .832
82 1.104
83 6.514
84 3.310
85 7.029
86 9.454
87 24.838
88 5.252
89 3.869
90 5.389
91 1.663
92 15.698
93 8.138
94 12.810
95 7.327
96 6.037
97 5.363
98 3.092
99 7.13?
14.511
10.262
11.347
25.628
59.482
5,983
5.460
13.044
15.433
7.529
20.703
9.364
2.068
12.775
10.904
18.073
1.148
2.111
7.297
26.067
10.801
.890
17.940
1.590
1.889
1.248
1.005
9.224
5.549
23.651
.451
1.329
14,302
13.140
.586
6.131
11.246
2.139
4.758
2.189
4.197
24.332
16.804
1.911
10.805
4.987
8.367
P.687
2.868
9.140
5.205
9.158
13.503
18.297
6.059
3.068
5.410
1.263
1 1 .040
4.726
10.011
4.040
4.048
8.086
4.885
9.921
13.098
6.900
11.208
19.743
50.343
5*024
4.882
9.567
13.835
10.606
24.415
9.369
.806
10.786
8.629
14.175
2.129
4.169
8.746
20.618
8.411
.354
15.845
.853
3.139
2.953
1.399
10.011
4.038
20.371
1.384
2.496
11.811
7.055
1.140
7.884
15.327
.673
5.905
3.776
5.030
22.008
15.286
3.328
10.266
4.808
9.268
2.387
2.956
9.164
4.375
9.892
12.003
20.625
4.311
4.287
5.497
1.681
12.985
8.852
11.505
5.726
4.321
6.740
2.802
8.738
9.087
6.790
8.387
16.386
44.161
8.562
8.238
16.107
19.100
18.641
37.013
11.769
4.274
9.215
8.644
13.124
3.942
5.4.63
6.861
17.507
7.345
1.611
12.013
1.925
2.966
2.684
2.381
6.729
2.643
15.633
1.466
4.576
10.005
6.621
2.598
12.721
19.771
4.603
5.886
3.627
4.620
17.032
12.481
5.434
8.500
2.566
6.386
2.408
1.867
5.812
3.020
6.856
9.035
32.871
10.028
8.084
10.270
4.295
18.559
13.217
13.067
7.435
9.321
4.282
3.705
9.945
3.352
.746
3.710
7.886
28.235
3*5.39
2.817
12.961
1L*140
19.324
39.846
4*112
3.238
2.538
^.036
3.729
9.602
._.8.9_07
3.161
7.625
1,438
4.716
4.901
6.628
9.885
3.069
2.471
2.554
1.015
7.410
4.361
10.750
2,367
4.449
7.146
21.229
28.733
6.610
4,341
3.227
1.857
8.504
4.861
12.488
2.479
1.230
3.030
1.145
1.445
1.905
1.733
3.037
2.932
34.964
5.167
5.116
4.752.
5.713
26.082
12.773_.
23.085
16.149
12.883
2.862
1.628
1.999
11.886 9.856 5.829
9.181 12.022 2.544
7.260 14.563 5.027
22.820 3.395 12.076
53.596 3.924 36.895
3.8J9 25_,2_8S 2,835_
3.756 25.067 2.242
13.011 41.286 11.732
JLZ.874 33,497 1 ljj?61
8.360 64.816 14.094
22.345 98.801 32.330
_6.2_41_ 22_».01Q_ 4 * 3_3.fl
2.717 31.298 1.589
11.191 10.671 4.295
J5.293 12.489 3,26L
15.030 6.530 6.885
2.359 44.735 5.602
1,115 4.1,.114 5j,455_
3.973 18.759 3.114
21.677 2.275 12.112
7.748 10.065 3.002
1.869 33.225 2.137
13.609 7.876 7.946
3.279 35.925 3.559
3.814 43.727 6.146
2.119 29.560 1.152
1.231 28.368 .620
6.181 13.366 3.216
5.833 19.115 1.434
18.523 2.730 11.073
1.146 32.700 1.796
2.811 46.475 6.265
11,264 8.630 4.938
12.469 13.326 5.751
1.944 39.762 3.733
8.200 64.857 14.662
9.960 73.305 21.374
3.460 38.048 3.784
3.429 24.115 2.609
1.445 26.533 1.316
2.037 21.005 1.198
20.096 9.386 12.799
11.647 8.442 7.297
4.437 51.186 7.798
7.189 13.181 3.889
3.854 16.658 .955
4.954 15.999 3.381
2.274 23.084 .433
2.327 22.190 .645
6.545 11.776 2.727
3.738 16.726 1.183
5.730 15.617 3.761
9.684 8.247 4.857
20.661 89.703 28.053
5.662 33.120 4.272
2.263 34.027 2.995
4.299 31.413 3.552
1.249 35.658 2.825
10.960 67.335 19.216
5.102 49.529 8.468
10.530 58.324 16.881
4.904 52.151 10.462
5.115 48.736 8,292
5.094 13.691 2.908
3.204 19.263 1.190
7.625 15.756 3.077
11.646
12.720
15.057
4.497
4.545
2J.859
24.163
38.324
30.. 127
63.137
95.095
5Q.771
31.343
12.100
13.228
6.697
46.000
4J.752
19.708
2.417
10.45Q
34.699
8.865
37.943
46.390
32.379
29.723
15.307
21.359
3.724
34.862
48.122
9.447
13.943
41.471
65.548
72.202
38.173
25.816
28.469
21.859
10.561
8.554
53.150
13.758
19.116
17.429
24.559
24.243
13.943
18.489
16.922
9.337
86.525
31.170
33.713
30.198
36.270
66.791
50.408
60.298
53.733
48.877
15.038
19.140
15.731
1.747
5.633
.654
11.406
27.159
9,406
8.260
24.562
17.135
29.094
53.247
6j.908
11.655
6.124
7,308
6.730
16.678
11.948
1.621
8.735
3.601
12.186
1.869
15.534
15.712
7.162
8.075
1.202
4.514
5.641
9.351
18.931
2.960
13.176
14.311
33.124
35.322
17.170
8.952
6.826
3.330
3.212
2.474
21.062
1.285
3.669
.999
4.788
4.198
1.667
5.835
.432
2.333
50,775
13.644
10.960
J1.312
12.593
37.310
19.545
32.441
24.972
24.397
4.697
7.251
5.570
1.324
2.440
2.374
5.157
16.303
8.959
7.972
21.935
15.393
33.650
59.026
6.788
11.726
3.798
4.623
2.205
21.245
17,562
4.616
3.074
J.888
14.325
.884
17.691
20.916
10.466
10.173
3.134
4.595
2.816
12.979
23.542
.736
8.398
17.967
38.607
44.422
17.801
11.063
9.981
5.611
1.907
1.788
26.077
1.761
5.059
3.663
6.199
6.285
2.777
6.487
2.926
1.651
54.403
12.280
12.943
11.409
15.223
42.979
24.483
38.478
30.817
26.894
5.437
6.717
3.899
8.980
11.361
13.857
2.884
3.855
24.181
24.362
37.736
29.632
63.232
96.407
20 . 75 1
30.659
9.859
12.034
5.898
44.680
41.640
18.526
1.795
10.175
33.577
8.173
36.663
43.946
30.308
28.180
13.694
19.039
3.050
33.244
47.336
8.849
13.107
40.242
65.634
73.274
37.695
25.231
27.436
20.917
8.540
8.404
51.273
13.193
17.605
16.345
23.377
22.802
12.263
17.601
15.473
8.667
87.306
31.119
33.675
30.350
36.325
68.894
50.264
61.425
53.341
49.609
14.878
18.691
15.410
9.888
12.064
17.293
3.374
2.551
28.047
27.685
39.634
31.451
68.483
102.886
23.954
33.400
10.950
13.988
7.001
49.925
48.155
22.850
3.161
13.334
37.836
10.622
41.255
48.362
33.344
31.650
17.465
20.040
6.185
37.473
53.402
10.635
15.412
44.947
73.635
84.543
41.657
30.757
32.250
25.083
8.662
12.193
56.115
16.422
21.318
20.746
26.299
26.160
15.726
22.209
19.154
12.152
92.880
33.022
37.830
33.415
41.661
79.730
55.942
72.080
61.924
56.315
20.528
22.735
17.639
3.137
1.397
3.950
8.345
29.920
5.243
5.025
11.413
11.212
20.762
40.935
6.516
3.630
3.661
3.873
5.431
8.467
9.243
4.489
8.678
3.032
4.371
6.142
6.115
8.040
4.321
2.916
4.191
.632
9.015
4.366
10.631
4.243
3.581
6.725
21.355
28.520
6.402
7.161
5.186
3.213
8.394
6.689
11.214
4.163
2.434
4.372
1.867
2.085
3.389
3.098
3.843
4.723
36.076
5.887
7.474
7.174
6.978
27.891
16.574
24.361
15.926
14.483
4.100
1.851
4.801
13.917
7.458
11.731
20.555
52.307
4.458
4.652
7.909
12.237
9.013
21.938
8.926
.662
11.011
8.762
14.730
1.875
4.243
9.008
21.492
9.027
.511
17.204
1.030
3.308
3.561
1.400
10.634
4.984
21.384
1.733
2.418
12.916
7.443
1.267
7.003
13.698
.309
6.005
4.090
5.209
23.488
16.003
3.071
11.020
5.598
9.925
3.032
3.717
9.891
4.811
10.518
12.849
18.098
3.616
3.936
4.978
1.732
11.936
8.410
11.535
5.167
3.783
7.419
2.634
9.011
235
-------
100
3.115 3.292 4.333 7.243 3.220 2.733 29.909 1.916
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
~T9
80
81
82
83
84
85
86
87
88
89
90
91
92
93
31
0
7.869
23.562
13.799"
10.415
11.950
26.260
60.109
8.581
8.417
9.740
13.711
10.260
22.629
12.421
2.715
14.595
13.936
19.871
1.843
4.804
10.356
27.230
13.588
2.260
19.456
3.429
1.650
4.287
2.860
13.046
4.760
26.955
2.361
4.206
16.445
12.976
2.073
10.511
16.137
3.065
11.704
7.171
7.283
22.304
19.918
2.555
13.327
8.554
11.671
4.220
4.888
12.797
9.585
11.127
17.016
20.502
5. 577
6.944
8.307
4.913
19.048
12.131
32
0
8.631
8.653
4.188
7.123
13.811
40.102
.497
1.056
5.695
3.517
10.004
25.545
1.420
2.490
4.475
3.114
7.208
8.875
9.132
4.659
13.502
5.014
5.998
12.352
8.192
11.602
5.721
2.642
6.041
6.399
13.779
6.347
10.499
8.130
8.369
8.068
16.732
20.705
5.096
4.532
4.474
2.789
16.997
9.023
11.632
6.371
S.~58T
6.641
4.405
5.459
6.074
4.476
6.665
7.743
20.815
1.354
2.369
1.169
5.604
20.313
8.627
33
0
4.699
4.955
5.637
4.560
16.765
8.706
8.580
23.920
16.008
34.256
60.763
6.056
14.856
2.507
2.745
2.020
24.935
21.290
5.333
3.129
2.492
17.353
4.581
20.008
25.986
12.969
11.882
2.830
10.184
1.659
16.166
25.842
3.214
9.482
21.621
38.396
43.078
20.253
7.437
9.658
6.421
8.722
2.227
29.899
3.980
5.807
4.912
10.071
9.679
2.489
5.284
5.232
1.726
53.978
15.468
13.382
11.966
16.959
39.529
22.468
34
0
2.860
1.950
5.681
20.147
9.691
9.211
19.783
13.931
32.399
57.791
7.543
11.511
2.882
4.922
3.616
18.902
17.164
4.089
4.794
3.376
13.265
2.654
16.354
17.620
9.556
8.890
2.787
3.738
4.320
11.603
22.214
2.782
8.469
16.560
37.080
42.215
17.108
11.430
9.842
5.367
2.261
3.406
23.110
2.955
4.921
3.687
6.267
6.030
2.521
6.602
2.451
2.875
52.815
12.540
13.560
12.395
15.733
43.608
24.958
35
0
4.931
4.327
21.905
4.776
4.419
12.549
10.404
24.236
46.414
4.935
5.396
1.177
1.172
1.724
14.161
14.336
5.284
4.627
1.582
8.004
5.132
10.040
14.491
6.576
5.058
3.916
2.359
5.834
8.193
15.864
2.280
3.028
11.450
27.300
36.075
9.357
6.889
6.537
4.139
7.770
4.956
17.422
3.685
3.162
5.163
3.705
4.192
2.917
3.299
4.818
3.190
39.971
6.331
8.232
6.688
9.748
32.738
17.977
36
0
11.169
28.087
6.751
6.256
19.543
12.510
25.204
47.349
4.739
9.761
5.672
6.438
6.022
14.998
10.589
1.073
8.234
2.969
11.096
2.330
14.625
15.038
7.591
7.069
1.509
4.753
5.829
8.934
17.665
3.106
11.796
13.293
30.535
31.296
14.886
8.576
6.655
2.536
3.835
2.014
19.653
1.022
4.108
1.021
4.502
4.305
2.105
5.564
.332
2.389
45.047
10.408
9.125
8.920
11.241
34.168
18.375
37
0
8.677
14.942
15.125
23.965
19.953
46.187
75.214
13.585
18.281
3.331
4.663
1.951
31.023
31.384
13.608
.996
5.640
21.318
7.622
23.331
30.644
19.383
17.053
9.427
10.716
3.342
21.915
33.095
5.566
5.115
27.116
47.971
58.728
23.556
16.059
17.954
13.767
9.154
7.457
35.631
9.768
10.464
12.401
14.561
14.519
7.626
9.989
11.817
6.000
65.398
19.177
22.414
19.424
24.533
53.071
36.182
38
0
40.292
39.451
57.598
47.057
88.681
127.257
35.041
48.266
20.437
22.900
13.456
68.397
64.058
34.200
7.450
21.160
53.177
17.127
56.763
67.268
47.335
46.346
26.932
32.839
10.278
52.834
70.533
17.466
24.950
61.646
92.768
105.053
57.723
41.089
44.480
37.054
16.668
18.716
75.855
25.225
31.714
30.821
38.975
38.587
24.812
32.158
29.910
19.080
116.049
48.097
51.563
46.586
55.718
95.868
70.726
39
0
.308
7.656
5.055
9.645
24.756
1.032
2.511
5.815
3.649
7.464
8.780
7.436
4.096
13.620
4.251
5.763
11.261
8.051
12.03F
5.522
2.830
5.669
6.952
13.200
6.052
9.664
7.326
9.046
7.670
15.616
18.774
5.047
3.517
3.586
2.216
17.171
7.875
11.918
5.349
5.114
5.716
3.886
4.918
5.831
3.877
6.181
6.891
20.870
1.607
1.455
.617
4.228
17.421
7.114
40
0
10.033
6.950
10.263
26.006
1.037
2.421
5.913
3.635
7.199
8.910
6.730
3.510
13.484
3.652
5.517
9.845
7.886
11.843
4.217
2.593
4.830
5.989
12.650
5.374
9.366
6.040
9.695
7.262
16.068
20.001
5.368
2.895
2.653
1.675
15.885
7.144
11.982
4.271
4.131
4.759
2.849
3.812
5.006
3.421
5.369
6.025
22.525
2.040
.965
.414
3.584
17.793
6.067
41
0
3.525
11.550
21.827
11.487
6.843
13.913
12.890
18.747
12.020
18.133
17.888
26.154
17.628
11.366
28.801
12.927
15.241
16.493
9.556
20.934
14.806
30.098
14.137
15.835
22.837
12.744
13.424
18.928
23.760
7.266
17.691
16.697
14.195
31.495
24.871
13.651
21.009
18.312
21.468
14.382
16.426
20.432
16.048
20.383
23.079
15.841
3.877
11.307
8.813
13.975
26.440
20.045
42
0
13.772
27.096
5.245
9.060
10.037
9.675
13.043
17.082
18.619
11.523
19.436
12.940
15.389
20.698
18.789
20.806
16.417
10.369
14.586
14.656
21.813
16.288
21.387
16.534
15.677
17.922
27.322
28.073
12.499
15.572
15.130
10.314
22.200
16.081
20.544
13.992
16.318
15.175
13.514
15.428
15.165
14.808
13.813
16.399
22.451
3.817
9.457
6.179
15.715
32.571
19.069
43
0
3.990
13.970
8.039
26.827
22.811
33.245
7.338
8.467
18.475
45.538
24.934
10.258
37.143
11.714
11.988
12.598
9.498
24.106
21.274
43.963
7.825
30.938
29.105
8.871
6.390
6.733
7.177
14.313
12.171
14.022
45.376
32.176
7.862
24.622
20.285
22.427
14.591
16.312
25.157
18.706
23.344
30.016
3.386
7.323
5.931
7.185
7.603
9.135
5.123
44
0
31.431
21.668
50.591
44.852
58.717
17.972
20.381
38.608
74.447
47.719
24.303
63.726
25.516
24.528
29.562
24.975
47.183
41.828
73.088
25.822
18.381
55.870
50.931
21.297
11.769
9.669
18.677
32.392
29.304
32.367
73.689
57.193
17.448
47.139
41.500
44.187
32.548
35.125
48.763
39.020
45.345
54.927
1.386
19.708
19.049
20.795
20.492
14.094
15.496
45
0
5.750
4.913
3.489
5.780
13.731
10.424
2.727
11.192
3.567
9.941
8.471
13.171
16.828
7.311
5.19S
4.013
8.088
10.197
9.187
15.111
5.536
11.893
12.235
23.166
24.867
10.059
4.735
4.972
2.242
13.137
5.072
17.737
3.503
5.772
4.309
5.277
6.191
4.630
5.147
4.451
5.055
27.893
3.787
3.060
1.697
7.734
24.553
9.608
236
-------
94
95
96
97
98
99
100
46
47
48
49
50
51
52
53
54
55
56
. 57
58
59
.. 60
61
62
63
64
65
66
"67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
18.327
8.876
9.327
12.600
6.445
_13.474._
6.350
.46
0
8.340
6.328
11.732
3.260
4.874_
7.022
18.827
7.161
1.307
14.861
2.451
4.945
2.678
.780
8.579
4.021
18.990
2.083
4.057
10.342
7.620
2.337
9.482
16.208
.886
4.923
3.221
3.662
20.394
13.462
4.725
8.679
4.755
8.256
2.153
3.089
8.081
4.251
8.630
10.749
17.744
1.636
2.166
2.467
1.883
14.475
6.910
15.092
7.842
4.818
7.173
2.029
5.944
2.098
23.317
13.957
9.681
7.112
1.663
2,881
2.247
47
0
,_636
1.354
17.764
IB. 119
5.549
4.002
2.409
11.297
6.371
13.403
18.203
8.857
6.915
3.658
5.001
4.845
11.146
19.708
3.481
4.551
15.202
31.201
39.026
12.731
7.118
7.942
5.110
8j829
4.952
21.257
4.732
4.393
5.845
6.455
6.681
2.811
4.104
5.460
3.346
43.073
8.747
10.110
8.238
13.060
36.313
19.773
34.044
24.549
20.744
5.650
3.725
2.1.810
7.894
36.262
29.600
25.929
4.800
7.130
3^708
11.328
. 0
1.492
15.776
15.509
5.258
4.932
1.621
9.357
6.941
11.213
17.228
7.495
5.584
3.675
5.314
5.535
9.619
16.642
3.248
~4.099
13.028
26.464
34.361
10.089
4.545
5.886
4.066
U«440_
4.900
19.169
4.371
3.555
5.568
5.307
5.732
2.818
2.649
5.830
3.129
37.650
6.799
7.234
5.624
9.853
29.800
15.432
28.382
20.837
15.930
4.230
2.493
1.416
5.737
38.993
28.817
27.420
6.380
6.436
10^148
_ 49
0
23.736
21.795
6.937
1.222
15.554
4.108
.iS.242
24.706
12.566
10^847
4.517
7.275
2.261
15.450
25.252
1.796
5.601
20.209
38.312
46. 149
17.216
9.348
10.717
7.004
...6j.928
3.039
28.367
4.083
6.067
6.489
8.560
8.917
3.623
5.744
6.276
2.274
51.355
11.454
12.844
10.317
16.221
41.529
24.013
38.708
30.973
25.219
5.874
5.777
2.652
10.220
29.926
21.671
17.382
4.533
2.512
2.328.
5.933
0
lK33f
31.879
1.251
23.377
.759
4.254
_2j_844
14.248
8.147
29.922
1.567
.765
19.861
14^921
.472
4.130
8.338
1.937
9.014
5.597
7.628
29.607
22.301
.417
15.634
8.805
12.665
5.339
5.631
14.080
8.719
13.235
18.879
16.102
7.516
5.953
8.630
2.619
9.884
8.057
9.402
2.705
4.337
11.632
6.792
1S.29J7_
6.834
31.409
23.151
22.132
4.445
5.315
4j_646
6.865
0
6.979
29.286
12.180
3.129
17.392
4.602
3.902
3.484
3j_60_6_
10.167
8.642
25.106
2.016
2.442
15.513
18.445
2.162
7,291
8.124
5.180
6.629
3.353
4.738
24.585
16.208
4.311
10.402
6.932
7.735
3.983
4.052
10.834
7.431
8.762
14.262
20.755
8.636
3.932
7.000
1.694
9.585
5.317
8.861
4.915
5.512
8.360
7.184
I2.»041
4.606
47.456
38.724
34.248
9.968
10.249
19.068
52
0
10.900
2*158
8.095
4.335
11.108
12.208
4.479
^29_T_
.758
4.876
7.685
5.987
12.794
3.865
12.170
9.696
23.210
23.999
11.245
4.267
3.022
.627
8.108
2.872
15.314
1.071
2.506
.379
2.907
2.738
1.469
3.347
.443
2.609
36.393
8.118
4.909
5.384
7.026
25.507
11.169
23.899
17.029
15.620
3.119
3.630
3.082
3.559
86.224
79.232
71.468
28.890
34.777
26.418
45.652
53
0
4.014
22.268
4.370
25.144
32.128
19.206
17.367
7.381
11.151
1.183
22.110
33.810
3.396
7.515
27.842
49.280
57.190
25.023
15.175
16.894
11.952
6.200
4.000
37.432
6.717
9.876
9.580
13.762
13.733
6.215
9.707
9.112
3.709
66.370
18.906
20.830
17.935
23.772
52.378
34.487
47.377
39.778
34.717
8.453
10.246
17.513
20.271 19.725 33.117 40.186 17.246 27.003
12.906 13.360 18.703 25.412 9.132 18.388
8.173 8.136 15.184 20.454 6.067 15.095
5.709 5.109 20.738 17.991 23.535 45.307
1.602 2.065 8.046 7.756 14.122 31.646
2.332 1.590 15.239 9.746 19.533 40.142
1.626 .735 12.161 9.490 8.612
54 55 56 57 58
0
9.070 0
2.448 16.316 0
11.460 .344 19.798 0
16.419 1.865 22.905 2.015 0
6.572 1.976 11.326 3.171 4.327
5.835 1.005 12.200 2.199 3.759
1.510 9.351 2.642 12.000 14.485
4.176 4.149 6.581 5.785 6.965
3.040 21.031 2.268 24.010 30.035
8.470 .502 14.135 1.287 1.729
16.002 1.452 24.886 1.277 1.968
.721 12.600 .979 15.515 20.178
5.523 8.496 11.757 8.897 15.132
12.300 .391 19.677 .673 1.037
27.369 6.721 40.570 5.711 7.092
33.074 13.535 44.807 13.916 12.555
11.426 .988 21.164 1.220 3.804
4.372 5.117 11.011 6.216 11.129
4.723 2.856 10.046 4.194 6.846
2.487 4.609 6.044 6.951 8.851
7.016 22.684 2.090 27.014 27.729
1.499 15.697 1.009 19.211 23.243
19.576 2.212 28.701 2.004 .879
1.133 10.406 1.399 13.633 16.327
1.973 4.256 5.174 5.675 8.777
2.175 8.612 2.743 11.459 13.005
3.572 2.236 7.225 3.927 5.615
3.641 2.473 6.799 4.035 5.515
1.045 8.667 2.578 10.952 14.094
1.923 4.057 6.896 5.146 9.594
2.552 9.363 2.489 12.505 13.221
.553 12.197 1.179 14.937 19.380
42.649 20.834 60.140 21.373 22.492
8.468 5.619 16.508 7.831 10.530
7.203 4.225 14.486 6.202 8.994
6.453 6.119 13.890 8.508 12.098
8.712 1.317 16.151 2.164 4.652
29.334 11.842 43.075 11.036 14.742
15.616 7.672 24.959 8.935 11.745
26.153 9.760 37.092 8.224 11.878
20.730 4.184 31.667 3.317 4.660
16.743 3.979 28.614 3.651 7.778
1.578 6.489 4.647 7.971 12.400
2.734 3.297 8.871 4.773 8.391
_L.L93 9.465 5.209 12.168 17.420
5.400 4.337 11.113 6.592 9.044
23.685
59
0
1.059
5.092
2.816
16.644
.815
4.219
8.412
11.071
2.407
11.976
18.460
4.431
3.213
.937
2.293
16.963
11.217
5.701
6.224
2.002
4.683
.843
.813
4.899
2.967
5.422
8.226
26.734
6.796
3.145
5.066
2.066
16.620
5.937
13.569
8.458
7.215
5.019
4.213
6.103
2.165
27.387
19.305
13.962
6.328
3.440
1.371
2.477
60
0
5.458
2.864
16.116
.846
3.652
8.754
7.955
1.870
10.084
15.612
1.992
3.152
1.420
1.850
17.460
10.917
4.475
6.508
2.565
5.245
1.089
1.432
5.195
2.535
5.646
8.280
21.353
3.383
2.119
3.119
1.677
14.937
6.337
14.082
7.013
5.715
4.951
1.660
5.290
1.786
62
71
237
-------
61 0
62 4.386
63 4.441
64 7.321
65 15.233
6ft 2.077
67 9.906
68 11.692
69 27.023
70 30.258
71 13.126
72 4.059
73 3.587
74 1.460
75 6.534
76 1.720
77 18.284
78 .757
79 1.564
80 .406
81 3.406
82 2.942
83 .185
84 2.403
85 .645
86 .936
87 43.477
88 10.632
89 7.149
90 7.478
91 8.865
92 29.437
93 14.004
94 25.725
95 19.669
96 17.968
97 2.013
98 4.218
99 2.588
100 5.212
76
76 0
77 27.843
78 .927
79 4.943
80 2.081
81 7.336
82 7.094
83 1.942
84 5.532
85 2.139
86 .530
87 53.334
88 13.869
89 11.550
90 10.631
91 14.456
92 37.928
93 21.137
94 34.481
95 28.789
96 25.246
97 3.638
98 6.859
99 3.252
0
10.841
3.501
9.916
4.726
5.735
6.000
21.250
29.567
6.976
7.320
4.630
3.505
8.762
8.071
10.373
4.584
2.104
4.470
1.308
1.357
3.604
3.583
4.048
5.589
37.480
7.348
7.673
8.039
6.715
28.511
15.532
23.729
16.000
14.704
4.877
3.493
5.641
5.249
77
0
20.196
11.582
16.799
7.674
7.990
17.903
11.517
17.457
23.607
15.306
9.744
8.204
11.280
4.135
10.210
9.272
9.578
2.764
4.781
15.319
9.616
19.162
0
19.753
31.116
2.394
9.947
25.845
46.538
52.006
25.178
12.160
13.697
9.431
5.244
1.737
35.680
4.306
7.561
6.061
13.043
11.461
3.809
7.798
6.087
1.687
66.580
30.240
18.886
17.312
21.309
47.922
30.488
41.976
36.285
33.700
5.690
10.056
6.260
15.831
78
0
3.571
.597
3.454
3.379
1.061
3.624
.678
.817
44.234
9.712
7.147
6.938
9.278
31.148
15.071
28.078
32.354
19.153
2.643
4.692
2.075
0
1.866
11.325
10.598
.579
8.233
13.805
2.533
4.541
1.814
3.315
19.972
13.858
2.763
8.495
3.243
6.376
1.422
1.399
7.041
3.784
7.046
10.730
33.307
6.585
3.815
6.113
1.352
13.231
6.798
10.539
5.081
5.397
5.788
3.888
8.812
3.518
79
"" 0"
1.841
1.124
.725
1.042
.511
2.572
2.624
37.193
8.703
5.377
6.489
4.602
22.093
10.812
17.321
12.784
11.361
1.022
2.519
3.297
0
20.726
16.078
.544
2.430
7.727
2.218
7.598
4.933
8.484
33.418
33.580
.977
16.788
8.823
13.678
5.931
6.089
14.883
8.344
15.102
19.596
16.396
9.418
5.839
9.157
1.714
6.387
6.039
5.188
1.271
2.184
11.175
7.927
15.741
7.123"
80
0
2.857
2.415
.873
2.958
.252
1.624
41.838
10.413
6.511
7.304
7.688
27.214
13.015
23.686
18.185
16.968
2.082
4.519
3.438
0
7.956
16.075
34.472
40.970
15.909
7.357
7.173
4.272
4.304
1.246
24.446
1.073
3.278"
2.745
4.979
1.642
4.184
3.837
.605
51.084
11.644
10.312
9.330
12.364
37.020
19.715
32.470
27.132
22.433
3.369
5.714
3.136
7.457
81
0
". 145
3.143
1.982
3.236
5.161
29.575
5.236
3.332
4.279
2.522
19.295
8.472
16.312
10.748
8.849
3.087
2.315
3.948
0
12.470
25.600
36.995'
9.089
10.085
'10.700
9.560
15.981
11.360
17.529
10.565
6.413
11.572
7.840
8.104
7.799
5.081
11.654"
8.236
42.243
10.464"
13.778
12.548
12."221
31.334
24.115
0
4.977
10.861"
1.546
6.647
3.660
6.019
26.381
19.139
1.049
12.860
6.149
10.403
3.376
3.571
11.276
6.234
11.255
15.515
18.940
7.116
4.824
7.444
"1.403
10.166
6.929
27.025 8.347
19.653 3.084
17.067 3.654
""6.539 87837""
3.887 5.528
8.300 12.388
12.387
82
0
2.663
1.727
2.881
4.812
32.166
6.894
4.123
5.562
2.926
19.745
9.161
15.766
10.632
9.569
2.540
3.743
4.564
5.232
83
0
1.724
1.176
.708
44.260
10.620
7.600
7.846
8.742
29.575
14.759
25.143
19.315
17.497
1.513
3.683
3.468
0
4.323
5.404
13.818
11.853
17.223
52.011
37.165
3.419
29.192
18.830
25.185
14.926
15.473
36.753
16.803
37.360
32.532
9.940
14.637
10.607
14.471
"5.984
2.494
7.833
4.486
1.266
2.148
20".94S'
15.247
25.629
13.852
84
0
4.013
3.105
33.931
7.799
4.767
5.608
4.089
18.981
9.946
15.346
11.123
9.234
.755
1.561
3.023
0
13.614
18.916
16.461
19.902
55.976
39.471
9.030
32.104
25.330
26.992
20.888
21.295
31.557
23.256
28.896
37.155
11.142
19.336
13.750
18.051
10.807
4.216
10.619
9.369
5.083
8.040
25.829
20.419
31.199
18.056
85
0
1.951
43.017
10.213
7.501
7.901
9.119
30.566
15.152
27.256
20.212
19.353
3.208
4.804
3.915
0
6.675
5.009
6.801
28.113
19.384
2.634
13.854
7.467
12.604
4.560
5.437
12.339
6.289
13.369
15.823
14.532
3.671
4.056
5.100
2.044 •
10.403
7.504
11.081
4.531
3.725
9.744
3.804
10.527
4.892
86
0
50.246
12.449
9.896
9.425
11.643
34.076
18.526
29.608
24.505
21.393
1.945
5.030
3.372
0
.877
2.297
19.716
8.201
11.554
5.648
2.262
4.512
3.514
3.409
3.851
1.259
6.243
5.758
28.067
8.038""
2.860
4.154
3.150
14.011
5.114
12.429
9.431
6.571
2.624
3.349
3.635
3.004
87
0
15.926
16.759
17.652
18.285
13.088
14.318
25.349
16.043
12.020
41.007
26.634
35.313
0
1.179
17.385
8.478
7.816
4.738
1.465
3.271
1.348
1.254
3.595
1.497
4.511
6.095
26.363
67699
1.964
3.724
1.570"
13.785
4.427
11.704
7.839
6.068
"" 2.832
3.065
4.239
1.681
88
0
3.730
1.401
5.369
19.322
9.053
23.563
13.859
8.691
10.975
3.366
5.713
0
10.958
4.474
10.851
1.843
1.433
1.138
1.339
1.309
1.791
1.769
1.530
3.368
29.607
5.236
2.444
3.061
3.700
19.756
7.713
18.577
12.332
10.350
2.410
1.946
2.336
1.581
89
0
.617
1.757
12.194
3.446
14.640
9.071
4.865
7.073
3.736
4.426
0
4.413
35.202
4.930
10.559
6.760
11.950
11.614
6.600
13.524
"5.093
5.254
70.124
20.854"
21.709
20.142
"24.511
57.991
35.412
51.712
42.250
40.002
11.316
13.985
10.264
16.987
90
0
3.833
16.434
4.775
20.112
13.052
7.332
8.127
3.465
3.198
238
-------
100
9.234 4.985 3.896 4.775 1.946 2.751 5.563 4.020 S.411 7.617 21.199 2.760 .452 .800
91
92
93
95
96
98
99
R
3
4
2
4
7
91
0
.031
.366
.125
.174
.064
.840
7
3
3
18
27
92
0
.562
.572
.OJ32
.793
• 277_
10
7
3
10
u
j
93
- 0_
.480
.774
.86_4_.
.081
.709
, 360
.997
94
0
3.017
%*6J7_
16.975
18.192
--£6^67.6
17.701
?
14
10
£0
11
95
.370
.000
.873
• T3£__
.621
96
0
12.595
8.629
14.749
7.204
97 98 99 100
0
2.965 0
_ 3^891 3.217 0
6.119 3.395 2.968 0
239
-------
APPENDIX F
LISTING OF THE MCKEON CLUSTER ANALYSIS PROGRAM
c
c
c
c
c
c
c
c
c
c
c
PROGRAM MCKEON
THIS PROGRAM OBTAINED FROM NATIONAL INSTITUTES OF HEALTH
COMPUTER CENTER IN 8ETHESDA, MARYLAND. MODIFIED TO RUN
UNDER OS3, AND CERTAIN CONVENIENCE CHANGES MADE
JUNE 1974 BY K. BYRAM, US ENVIRONMENTAL PROTECTION AGENCY
MCKEON CLUSTER ANALYSIS VERSION I
DIMENSION B(150, 150), LV(150, 150),
1 KV(150), KC(150), C(150), FMTU80)
2 ,R(150),H(150),IOUT(150)
INTEGER C,FMT,HARDWARE
REAL1 B
EQUIVALENCE (B,LV),(R,C>,(H,KV),(FMT,IOUT)
PROGRAM READS UNIT 60
WRITE UNIT 61
DO 1 LUN=1,63
IF(.NOT.HARDWARE(LUN))GO TO 3
1 CONTINUE
STOP 1
3 CALL EOUIP(LUN,5HFILE )
INPUT FROM USER
C....SPECIFICATIONS INPUT
2 READ(60,2020)N,KOP,NV,KRIT1,KRIT2,MMCOL,MATOP
IF(EOF<60))GO TO 216
2020 FORMAT (1014)
IF(N.L£.0)STOP
NNCOL=20*MMCOL
READ(60,1199) (FMT(II),II=1,NNCOL)
1199 FORMAT (20A4)
WRITE(61,1198)N,KOP,NV,KRIT1,KRIT2,MMCOL,MATOP
MATOP=l+MATOP
1198 FORMAT( 1MCKEON CLUSTER ANALYSIS VERSION I.I /
» OCONTROL INPUT: ,1014)
WRITE(61,1197) (FMT(11)»1I = 1,NNCOL)
1197 FORMAT(14X,1H:,20A4)
WRITE<61,1196)
1196 FORMAT( OFIRST ROW OF DATA FOR VERIFICATION- )
NM1=N-1
FN=N
GO TO (5,18,7,7),KOP
KOP=1
C
C.
C • • • •
C
5 DO
1195
6
INPUT DATA
TRIANGULAR
6 1=1, NM1
AS CORRELATION MATRIX IN UPPER
FORM, MINUS DIAGONALS
READ(60,FMT) (6 ( I , J) , J=IP1 ,N)
IF ( I.EQ. 1 ) WRITE (61 » 1195) (B ( I , J) , J=IP1 ,N)
FORMATHOE12.3)
CONTINUE 240
00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00022
00023
00024
00025
00026
00027
00028
00029
00030
00031
00032
00033
00034
00035
00036
00037
00038
00039
00040
00041
00042
00043
00044
00045
00046
00047
00048
00049
00050
00051
00052
00053
00054
00055
00056
-------
c
c
c
B(N»N)=1.
IFCNV.NE.DGO TO 25
REFLECTION IF REQUESTED
WRITE<61»802>
802 FORMAT(1H »19HVARIABLES REFLECTED//)
DO 804 I=1»N
DO 804 J=I»N
804 B(J»I)=B(I»J)
WRITE(LUN) «B(I.J)»J=1»N).I = 1,N)
REMIND 2
DO 856 J=1»N
B(J»J)=0.
KC(J)=1.
BM=0.
DO 854 I=1»N
TEMP=A8S (8(I»J)>
IF(TEMP.LE.8M)GO TO 854
BM=TEMP
H(J)=8(I»J)
IM=I
854 B(I»J)=0.
856 B(IM,J)=H(J)
DO 810 1=1»N
R(I)=0.
DO 810 J=1»N
810 R(I)=R(I)*B(I»J)
ITR=0
811 RM=FN
DO 816 1=1«N
TEMP=R(I)+H(I)
IF(TEMP.GE.RM)GO TO 816
RM=TEMP
IM=I
816 CONTINUE
IF(RM.GE.O.)GO TO 890
DO 822 J=1»N
B(IM»J)=-B(IM»J)
822 B(J»IM)=-8(J.IM>
DO 824 J=1,N
H(J)=H(J)+2.»B(IM»J)
824 R(J)=R(J)+2.*8(J»IM)
H(IM)=-H(IM)
R IM
830 FORMAT (1H »6X»I8)
ITR=ITR+1
IF(ITR.LT.N)GO TO 811
890 CONTINUE
READ(LUN) «8
-------
18 00 20 1 = 1 »N
READ(60»FMT) (B(I»J)»J=I»N)
IF
20 CONTINUE
GO TO 25
C
C.
C
,.KOP=3 OR 4.
.INPUT DATA AS RAW NUMBERS
7 00 8 1=1, N
READ(60»FMT> (8 ( I » J) » J=l »NV)
IF(I.EQ.1)WRITE<61»1195) <8(I»J),J=1»NV)
8 CONTINUE
IF(KOP.EQ.4)GO TO 602
C...KOP=3... STANDARDIZE OVER COLUMNS
DO 10 J=1»NV
SUM=0.
ss=o.
DO 9 I = 1,N
SUM=SUM+B(I»J>
9 SS=SS+B»*2
FMEAN=SUM/FN
FMSQ=SS/FN
SD=SQRT (FMSO-FMEAN*»2)
DO 10 I=1»N
10 B(I»J)=-FMEAN)/SD
C... .COMPUTE DISTANCES. ...USING
602 IF(N.GT.75)GO TO 14
C ......... TOP OF MATRIX SINCE N LT 75
DO 12 1 = 1, N
DO 12 J=ItN
B( I+75» J) =0.
00 12 K=ltNV
12 B(I+75»J)=B(I+75»J)+B(I»K)*B(J»K)
DO 13 1=1. N
DO 13 J=I,N
13 8(ItJ)=8(I*75»J)
GO TO ?5
C. ........ USING EXTERNAL SCRATCH FILE SINCE N GT 75
14 DO 15 J=1,NV
15 WRITE(LUN)
DO 17 1=1, N
DO 17 J=I,N
17 B(I,J)=B(I,J)+R(I)*R(J)
INPUT COMPLETE. ..INITIALIZE
25 TB=0.
DO 26 1=1, NM1
IP1=I*1
DO 26 J=IP1,N
B(I,J)=B
-------
00 28 J=IP1»N
28 B(J»I)=BU»J)
NRC=N
KSH=0
LOC=1
IF(KRIT1.EQ.N.AND.MATOP.EQ.1)GO TO 1021
WRITE<61,2000)
2000 FORMAT ( -SQUARED DISTANCES BETWEEN POINTS BEFORE CLUSTERING )
GO TO 1000
1021 CONTINUE
IF(MATOP.EQ.3)WRITE<61»2014)
NC=N
NCM1=NC-1
BMIN=0.
TEMP=SORT (FN)
IF(KRITl.LE.O)KRITl=2.*TEMP+.5
IF(KRIT2.LE.O)KRIT2=2
MCL=0
MIS=N
DO 120 I=1»N
120
KV(I)=I
KC(I)=I
MAIN LOOP
C...FIND MINIMUM ENTRY IN OFF DIAGONAL ELEMENTS
130 BPR=8MIN
BMIN=1.E30
DO 134 I=1»NCM1
134
135
420
139
136
DO 134 J=IP1,NC
IF(B(I»J) ,GE.BMIN)GO TO 134
BMIN =B(I»J)
IM=I
JM = J
CONTINUE
DEL=BMIN-BPR
IF(NC.GT.KRIT1)GO TO 139
GO TO (149»1022»135)»MATOP
WRITE<61»420> BMIN» IM, JM,DEL»MIS»MCL
FORMAT <1H » 8X,F8.3, 9X » I3» 1H» » I3» 8X.F10.3t 18X. I8» 15X, I 10 )
CN=C(IM) +C( JM)
DO 136 K=1»NC
B(IMtK) =AMAX1 (B(IM»K)»6
-------
C...ADJUST NUMBER OF ISOLATED POINTS
IF(C(IM),LE.1)KTR=KTR+1
IF (CUM) ,LE.1)KTR=KTR+1
GO TO (311,314,313), KTR
311 MCL=MCL-1
GO TO 314
313 MCL=MCL*1
314 MIS=NCM1-MCL
IF(JM.GE.NC)GO TO 321
c...CLOSE UP COUNTER ARRAYS
DO 148 J=JM»NCM1
C(J)=C(J*1)
148 KC(J)=KC(J+1)
321 C(IM)=CN
NC=NCM1
NCM1=NC-1
GO TO 130
149 WRITE<61»2006)
2006 FORMAT(1H1 ,44HMAXIMUM SQUARED DISTANCE BETWEEN CLUSTERS »
1 53HMAXIMUM SQUARED DISTANCE WITHIN CLUSTERS IN DIAGONALS)
NRC=NC
KSH=75
LOC= 2
C LOWER TRIANGULAR MATRIX PRINTOUT
1000 DO 1002 JL=1»NRC,15
JU=JL+14
IF(JU.GT.NRC)JU=NRC
WRITE<61,1010>(J.J=JL»JU)
1010 FORMAT(1HO»2X,15I8)
DO 1014 I=JL,NRC
JUU=JU
IF(JUU.GT.I)JUU=I
1014 WRITE(61,1016)I,
WRITE(61,2014)
2014 FORMAT(1HO »7X, 22HNEXT CLUSTER. MAXIMUM ,11X,
1 17HCHANGE IN MAXIMUM ,11X ,9HNUMBER OF
2 14X,17HCLUSTERS WITH TWO/lH ,7X,26HSQUARED DISTANCE,
3 16HSQUARED DISTANCE,12X ;15HISOLATED POINTS,8X,
4 14HOR MORE POINTS//)
IF(NC.GT.KRIT2)GO TO 135
WRITE(6l,420) BMIN, IM, JM, DEL,MIS,MCL
c....RETURN TO GET ANOTHER CASE OR EXIT
GO TO 2
216 CALL UNEQUIP(LUN)
END
LOCATION,
00241
00242
00243
00244
00245
00246
00247
00248
00249
00250
00251
00252
00253
00254
00255
00256
00257
00258
00259
00260
00261
00262
00263
00264
00265
00266
00267
00268
00269
00270
00271
00272
00273
00274
00275
00276
00277
00278
00279
00280
00281
00282
00283
00284
00285
00286
00287
00288
00289
00290
00291
00292
00293
00294
00295
00296
00297
00298
00299
00300
00301
00302
00303
00304
244
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO. 2
EPA-600/3-76-037
4. TITLE AND SUBTITLE
Trophic Classification of Lakes Using LANDSAT-1
(ERTS-1) Mul ti spectral Scanner Data
7. AUTHOR(S)
D. H. P. Boland
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Special Studies Branch
Corvallis Environmental Research Laboratory
Corvallis, Oregon 97330
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
3. RECIPIENT'S ACCESSI ON> NO.
5. REPORT DATE
April 1976
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO.
10. PROGRAM ELEMENT NO.
1BA608
11. CONTRACT/GRANT NO.
13. TYPE OF REPORT AND PERIOD COVERED
Final - 1972-73
14. SPONSORING AGENCY CODE
EPA/ORD
15. SUPPLEMENTARY NOTES
Prepared in cooperation with
the National Aeronautics and
llt.i 1 i 7atinn. Wa^hinntnn. DC.
the Jet Propulsion Laboratory, Pasadena, California and
Space Administration, Office of Technology and
16. ABSTRACT
This study evaluates the Earth Resources Technology Satellite One (ERTS-l; j_.e_., LAND-
SAT-1) multispectral scanner (MSS) as a means of estimating lacustrine trophic state.
Numerical classificatory methods are employed to ascertain the trophic character of 100
lakes in Minnesota, Wisconsin, Michigan, and New York. Principal components analysis
is used to derive a multivariate trophic state index (PCI) using the trophic indicators
chlorophyll a_, conductivity, inverse of Secchi depth, total phosphorus, an algal assay
yield, and total organic nitrogen. A binary masking technique is used to extract lake-
related MSS data from digital tapes (CCTs). MSS color ratio models are developed which
give good estimates of Secchi depth and fair estimates of chlorophyll a_ levels. Tro-
phic state, as defined by lake position on the first principal component axis (PCI), is
predicted using MSS color ratio regression models. Each date of LANDSAT-1 coverage has
its unique model. An automatic image processing technique is employed to classify a
group of Wisconsin lakes. The utility of the LANDSAT-1 MSS is most apparent when the
seasonal contrasts between lakes at different points on the trophic scale are at a
maximum. Periods of excessive cloud cover, frames with faulty or missing MSS data, an
.
the need for some ground truth, impair, but do not preclude its use in lake monitoring
and classification. The use of CCTs in conjunction with digital image processing tech-
niques is essential if the maximum benefits are to be derived from the LANDSAT-1 MSS.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
*Remote sensing, *Lakes, *Multivariate
analysis, Artificial satellites, Water
quality, Spaceborne photography, Taxonomy
13. DISTRIBUTION STATEMENT
Release Unlimited
b. IDENTIFIERS/OPEN ENDED TERMS
ERTS-1 , LANDSAT-1 , Lake
classification, Trophic
state index, Multtspec-
tral scanner, Eutrophi-
cation, Automatic image
processing, Minnesota,
Wisconsin, New York, MI
19. SECURITY CLASS (This Report)
Unclassified
20. SECURITY CLASS (This page)
llnrla^ifiprl
c. cos ATI Field/Group
08/H, 20/F, 22/B
21. NO. OF PAGES
263
22. PRiCE
EPA Form 2220-1 (9-73)
245
U. S. GOVERNMENT PRINTING OFFICE 1976—697.057182 REGION 10
------- |