United States
Environmental Protection
Agency
Office of Research and
Development
Washington, D.C. 20460
EPA/600/R-00/012
July 2000
A Research Plan for the Use
of Thermal AVHRR Imagery
to Study Annual and
Seasonal Mean Surface
Temperatures for Large
Lakes in North America
322LEBOO.RPT •> 9/18/00
-------
EPA/600/R-00/012
July 2000
A Research Plan for the Use of
Thermal AVHRR Imagery
to Study Annual and
Seasonal Mean Surface Temperatures
for Large Lakes in North America
Principal Investigator
S. Taylor Jarnagin
Co-Investigator
E. Terrence Slonecker
NERL/ESD/LEB/EPIC
12201 Sunrise Valley Drive
555 National Center
Reston, Virginia 20192
National Exposure Research Laboratory
Ecological Exposure Research Program
Research Task Area Plan
-------
Table of Contents
Page
Abstract v
Task Summary v
Section 1-Goals and Objectives 1
Section 2-Background/Literature Review 2
Proposed Research 14
Section 3-Technical Approach 23
3.1 Work Breakdown Structure 23
3.2 Potential Study Areas 26
3.3 Results and Reporting 26
Section 4-The Importance of this Research to EPA Milestones and Program Area Goals 27
Section 5-Quality Assurance Statement 29
Section 6-Anticipated Results/Specific Work Products and Dates for All Sub-tasks 30
Section 7-Key Milestones 31
References 32
Appendix 39
-------
List of Figures
Figure 1. 1855 -1998 Lake Mendota WI ice cover duration 3
Figure 2. 1888 -1999 Houghton/Calumet seasonal snowfall amounts 4
Figure 3. Lake-wide average surface and volume mean temperature in Lake Superior 7
Figure 4. Seasonal cycle of the lake-wide averaged vertical distribution of temperature (°C) in
Lake Superior 8
Figure 5. Thermal structure of the lower Great Lakes 9
Figure 6. Simulated temperature fields under present and possible future conditions 10
Figure 7. Hypothetical 10°C isotherm depth with climate warming 11
Figure 8. Summer 1995 Lake Michigamme temperature 12
Figure 9. 1995 -1997 Lake Michigamme surface vs. whole depth mean temperatures 13
Figure 10. 1995 -1997 Lake Michigamme surface vs. epilimnetic temperatures 13
Figure 11. 1995 -1997 Central Basin Lake Michigamme mean temperature of the whole
water column 13
Figure 12. 1995 -1997 Central Basin Lake Michigamme temperature of the water surface 14
Figure 13. Summer 1995 -1997 Lake Michigamme temperatures 14
Figure 14. Annual average secchi depth (± 1 s.d.) in Lake Tahoe 16
Figure 15. Lake Superior mid-lake moored buoy locations 17
Figure 16. Lake Superior NDBC buoy 1994 water temp 17
Figure 17. 1989 -1998 Lake Superior buoy 45001 water temperature 19
Figure 18. 1987-88 to 1998-99 seasonal snowfall vs. prior October water temperature at buoy 45001 19
Figure 19. 1994 AVHRR CoastWatch MCSST pixel vs. buoy temperature 20
Figure 20. 1994 Lake Superior buoy - pixel temperatures 20
Figure A1. AVHRR thermal image (from CoastWatch MCSST data) of Lake Superior in
mid-August 1994 39
-------
List of Tables
Table Al. A listing of the 11 GCFs used 40
IV
-------
Abstract
Title: The use of thermal AVHRR imagery to study annual and seasonal mean surface temperatures for
large lakes in North America, the relationship between annual and seasonal surface and whole lake mean
temperatures, and the impact lake surface temperature changes have upon the surrounding environment.
Program Area/Task Area; Objective/Sub-objective: GPRA Goal 6.2.3: Reduction of Global and Cross-
Border Environmental Risks; Global Climate Change.
Principal Investigator: S. Taylor Jarnagin
Co-Investigator: E. Terrence Slonecker
Organization: US EPA Environmental Photographic Interpretation Center (EPIC), Reston VA.
(EPA/ORD/NERL/ESD-LV/LEB/EPIC)
Task Summary
An indicator is a value computed by statistically combining and summarizing data from a measurable
variable that yields information about larger processes that are more difficult (or impossible) to directly
measure. Ecological indicators are based upon measurements made of environmental parameters that can
be used to help understand ecological conditions or processes. This research proposes to use annual and
seasonal measurements of the whole-lake surface temperatures of large lakes as a landscape-scale
ecological indicator of the overall thermal content of those lakes. This research will investigate the use
of annual and seasonal values of whole-lake mean surface temperatures as a new, integrative indicator of
the ecological condition, integrity, and sustainability of large lake ecosystems that can be incorporated
into long-term monitoring programs.
Changes in lake thermal values may provide direct evidence of climate change. Tracking this
research plan's thermal indicator over time could reveal changes in the thermal content of lakes related to
changes in meteorological conditions and thereby act as an indicator of climate change. Establishing a
current relationship between the thermal characteristics of large lakes and the snowfall received in their
meteorological lee could allow for the use of historical snowfall records to make inferences about
historical changes in the thermal characteristics of large lakes. Changes in the estimator values will be
compared with meteorological variations and with other indirect measures of changes in regional climatic
conditions (e.g. ice cover duration and lake-effect snowfall). The process being developed through this
proposal could have wide applicability. Seasonal thermal estimates could be computed for a wide range
of lakes large enough to be sampled at the Advanced Very High Resolution Radiometer (AVHRR) pixel
size. The period of record for AVHRR imagery runs from 1979 to present. While the proposed research
focuses upon National Oceanic and Atmospheric Administration (NOAA) CoastWatch imagery for the
period 1992-1999, retrospective AVHRR imagery for periods prior to 1992 is available through the
National Environmental Satellite, Data, and Information Service (NESDIS) Satellite Active Archive
(SAA) and other providers. Future research could use additional AVHRR imagery to extend the time
frame of this study.
-------
This research will use currently existing, public domain data and imagery gathered by Federal
agencies and private groups. This will reduce the cost of the research and increase the utilization of data
products that have already been created at taxpayer expense. The research consists of three main tasks:
1) compare different methods of using remotely sensed satellite imagery to compute daily, seasonal,
and annual values of whole-lake surface temperature of large lakes:
This research will develop and test a method of computing the surface temperatures of large
lakes through the use of remotely sensed AVHRR thermal imagery and compare that method
with the NOAA Great Lakes Environmental Research Laboratory (GLERL) Great Lakes Surface
Environmental Analysis (GLSEA) method. NOAA CoastWatch imagery will be obtained via the
NOAA CoastWatch Archive and Access System (NCAAS) and used to compute a seasonal
estimate of the surface temperature of Lake Superior for 1994. The image acquisition season
will cover the period from late April until late October. The 1994 time series of images will be
processed to replace cloud-impacted or missing thermal values with temporally interpolated
values. The surface temperature of the lake recorded at each thermal image lake pixel will be
summed for each date to provide a mean surface temperature of the entire lake on the image date.
The seasonal series of heat content estimates will be integrated over the season to compute an
estimate of the seasonal surface temperature of Lake Superior. The seasonal surface temperature
estimate will be compared with analogous values calculated from the buoy data and with
seasonal values computed from the average temperatures calculated with the GLERL GLSEA for
1994. This comparison will allow for an assessment of the comparability of the two methods of
estimating mean surface temperatures of the Great Lakes. Either or both methods could be
applied to look at surface temperature values of many lakes large enough to be imaged at the
resolution of AVHRR imagery.
2) investigate the relationship between the surface temperature of Lake Superior and the impact the
lake has upon its surrounding environment in the form of lake-effect snowfall:
GLSEA values will be used to compute seasonal and annual mean surface temperature estimates
for Lake Superior for the period 1992-1999. Ground-truth for the temperature values calculated
from imagery will be obtained from NOAA National Data Buoy Center (NDBC) moored and C-
MAN buoys. The "thermal indicator season" as defined for this research will start on the latest
date that moored buoys were deployed and end on the earliest buoy removal date during the
period 1992-1999. This will allow for a uniform time series of values to be compared between
years. The seasonal temperature values of the lake will be compared to seasonal snowfall from a
series of stations located in the meteorological "lee" of Lake Superior. The relationship between
the surface temperature of the lake and the amount of snowfall will be statistically explored.
Additional geographic (e.g.: latitude, altitude, etc.) and climatic data (i.e.: air temperature, solar
insolation, ENSO cycle, etc.) will also be assessed on a regional basis to see if the addition of
other spatial and climatic variables helps to explain the observed variability in the amounts of
snowfall received in the lee of the Great Lakes. A correlation analysis among all five Great
Lakes will be done to see if the yearly temperature differences among lakes follow similar
seasonal and yearly trends.
3) investigate the relationship between surface temperatures of selected large lakes and the
temperature of the entire water column for those lakes:
VI
-------
Surface and vertical temperature data will be obtained from several large lakes with surface areas
large enough to be effectively sampled with AVHRR imagery. Yearly and seasonal patterns of
surface and whole water column thermal values will be compared to estimates of surface
temperature obtained from NOAA CoastWatch imagery for those lakes for the period of time
where concurrent imagery and in situ data exist. The lakes chosen for this study will include
lakes with a varying surface area to depth ratios, salinities, and geographic locations. Large
shallow lakes are relatively well mixed and are more likely to be isothermal over their depth.
Lakes with relatively smaller surface area to depth ratios (very deep lakes) are expected to be
highly stratified and heterogeneous in temperature. Salinity and geographic location are also
important factors in determining the thermal regime of a lake. Examples of large shallow lakes
include Oneida Lake, New York; Lake Okeechobee, Florida; and the Salton Sea, California.
Examples of very deep lakes include Pyramid Lake, Nevada and Lake Tahoe, California. The
Salton Sea and Pyramid Lake are examples of high salinity systems. The comparison of seasonal
values of surface temperature derived remotely from AVHRR imagery with surface and whole
water column values will allow for an assessment of the ability of AVHRR imagery to accurately
estimate changes in lake thermal values over time.
VII
-------
Section 1
Goals and Objectives
The goals of this research are to use thermal AVHRR imagery to study differences in the seasonal
mean surface temperature of Lake Superior, see how those differences in lake surface temperature relate
to the environmental influence Lake Superior has on its surrounding area, compare techniques and
methods of obtaining seasonal mean surface temperature values, and to study the relationship between
seasonal measurements of surface and whole water column temperatures. If this research is successful, a
single numeric estimator derived from thermal AVHRR imagery could be used as an indicator of the heat
content of large lakes. Changes in the seasonal heat content of lakes over time could be tracked remotely
over time with this indicator. The AVHRR database covers 1979 to the present and this process could be
applied to a wide range of lakes large enough to be sampled at the AVHRR pixel size. Archival AVHRR
data are available through the NOAA National Operational Hydrological Remote Sensing Center
(NOHRSC), the GLERL CoastWatch Program, and the National Environmental Satellite, Data, and
Information Service (NESDIS). AVHRR products for the period from 1985 to the present utilize Local
Area Coverage (LAC, 1 km nominal spatial resolution) AVHRR data, but NOAA only archived Global
Area Coverage (GAC, 4 km nominal spatial resolution) data prior to 1985.
Seasonal changes in the heat content of large lakes have direct ecological and economic impact if
those changes result in changes in fish and invertebrate communities or algae and aquatic vegetation.
Changes in the extent and duration of ice cover affect shipping and navigation on the Great Lakes.
Changes in the heat content of large lakes could alter the local climatic effects of those lakes on their
surroundings (such as seasonal lake-effect snowfall amounts). Changes in the seasonal heat content of the
epilimnion of large lakes tracked over time may reveal trends due to climate change. Seasonal changes in
the heat content of large lakes may be correlated with changes in other ecological phenomena that are
currently being suggested as indicators of climate change. These indirect climate change indicators
include such phenological measurements as the duration of ice cover on lakes, terrestrial snow cover
duration, spring snowmelt runoff characteristics, and blooming dates for flowering plants.
The first milestone in this research project is to obtain GLSEA estimates of surface temperature for
Lake Superior, temperature data from the NOAA National Data Buoy Center (NDBC) moored and C-
MAN buoys, and snowfall amounts for weather stations in the lee of Lake Superior for 1992-1999. The
second milestone will be to statistically compare the relationship between lake surface temperature,
snowfall, and other meteorological data available for those years. The third milestone will be to create
the standard operating procedure (SOP) to generate the estimator of seasonal large lake water
temperature and generate this estimator for Lake Superior for 1994. The fourth milestone will be to
compute this estimator and compare the values generated with those from the National Oceanic and
Atmospheric Administration (NOAA) Great Lakes Environmental Research Laboratory (GLERL) Great
Lakes Surface Environmental Analysis (GLSEA) method. The fifth milestone will be to obtain in situ
surface and vertical temperature data for large lakes viewable with CoastWatch AVHRR imagery and
concurrent imagery for those lakes. The sixth milestone will be to compare the remotely sensed
temperatures with the in situ measurements of surface temperature and whole water column
temperatures.
1
-------
Section 2
Background/Literature Review
The consequence of climate change and global warming on lakes has been the subject of much
concern and research (see DeStasio et al, 1996 for a partial review and listing). Globally, the impacts of
climate change are predicted to be most noticeable at higher latitudes (e.g. Mitchell et al, 1990;
Chapman and Walsh, 1993; Groisman et al, 1994). The Great Lakes of North America form the world's
largest reservoir of freshwater and are a major economic and ecological resource. Recent research has
focused upon how the biotic and abiotic environments of these lakes and their surroundings may be
affected by climate change (Croley, 1990; Hobbs et al, 1994; Croley et al, 1995; Mortsch and Quinn,
1996; Changnon, 1997; Mortsch et al, 1997; and Quinn, 1998).
Lakes, particularly large lakes, have a significant impact upon the local climate of the land adjacent
to the lakes due to the differences in heat capacities between the water and land surface and the moisture
supplied to the lower atmosphere by the lakes. The Great Lakes of North America moderate maximum
and minimum temperatures of the region in all seasons, increase cloud cover and precipitation over and
just downwind of the lakes during winter, and decrease summertime convective clouds and rainfall over
the lakes (Phillips, 1978; Scott and Huff, 1996). Perhaps the most dramatic examples of local climatic
impact of large lakes are the snow-belts extending downwind of the North American Great Lakes.
Locally, up to 100 - 300 % more precipitation falls downwind from the Great Lakes in winter than would
be expected if not for the influence of the lakes (Rothrock, 1969; Scott and Huff, 1996). The increased
snowfall in lake-effect areas is the result of a complex interaction of variables. Lake-effect snow is in
part a function of the heat and moisture lost from the relatively warm lake to the colder air mass moving
over it, the temperature of the air at the surface and at altitude, the temperature difference between the
lake and adjacent land, orographic precipitation due to air mass lift with the difference in terrain
elevation between lake and the adjacent land, and pressure gradient effects due to movements of air
masses (Jiusto, 1973; Dockus, 1985; Niziol 1987, 1989; Niziol et al, 1995).
Predicting synoptic-scale snowfall in mid-continental North America can be achieved with surprising
accuracy knowing only a single variable (net 12-hour vertical displacement, Chaston 1989). In contrast,
lake-effect snowstorms occur at a mesoscale (Orlanski 1975) and are complex-interaction events.
Rothrock (1969) found that the best single parameter correlating with significant lake snows was the
temperature difference between lake water and air at 850 mb. Thirteen ° C approximately equals the dry
lapse rate in this layer and 97 % of the cases of significant snow occurred when the water temperature
exceeded the air temperature by 13° or more. Jiusto and Kaplan (1972) noted that lake-effect snowfall
was heaviest when the body of water was warmer, increasing the vertical flux of momentum, heat, and
moisture over the water. They also noted that areas under upper-level support (positive vorticity
advection) were subject to greater snowfall.
Niziol (1987) identified the temperature difference between the surface of the water and 850 mb as
one of the "more important" forecast variables and placed this parameter at the start of his decision tree
for forecasting lake-effect snow. Niziol (1989) discusses a major single banded mesoscale lake-effect
snow to the lee of Lake Ontario. Terrestrial conditions predictive of lake-effect snow are discussed in
-------
detail but water temperatures of the lake are referred to only as being "relatively warm". Assuming the
water temperate is = 2.2° C, Dockus (1985) noted that 850 mb temperature = -10° C and a fetch of 160
km or more over open water was needed for true lake-effect snow (as compared to "lake-enhanced"
snow). Dockus also noted that upper-level support (positive vorticity advection) enhanced snowfall
amounts and reduced the oven-water fetch needed. It is apparent that knowing only the temperature of a
lake will not allow for a prediction of lake-effect snow. However, temperature differences over time,
interacting with other variables, may be related to changes in lake effect snowfall over time.
Lakes integrate regional
changes and act as indicators
of watershed conditions and
climate change. Long-term
records of ice cover exist for
some lakes and there are
observable trends in the ice
cover data over time
(Robertson, 1989; Magnuson,
1990; Magnuson et al., in
press). The reduced duration
of ice cover over time is an
indicator that the climate is
warming in the mid-
continental region of North
America over the period
represented by the data.
Figure 1 displays the
reduction in the duration of
ice cover on Lake Mendota,
Wisconsin over the past 144
years.
o
18
175
150
« 125
o>
o>
Q.
I
u
0
ti
HIM
O
ss,
ra
Q
100
75
50
25
y = -0.1583x
R2 =0.1426
410.1
185018601870 188018901900 1910 1920 19301940 1950 1960 1970 1980 19902000
Year at Start of Winter Season
Figure 1. 1855-1998 Lake Mendota Wl ice cover duration. Ice-cover
duration on Lake Mendota, Wl (data from the Lake Ice Analysis
Group (LIAG), Center for Limnology, UW-Madison). The Lake
Mendota ice cover data set is one of the longest sequential
phenology data sets used to search for indications of climate
change. The shorter duration of lake ice cover over time is an
indication of climate change (warming) over time.
There are three periods
within the data set with
apparent stepped reductions
in ice cover duration (1855 - 1886, 1887 - 1978, and 1979 - 1997). There is a significant difference in ice
cover duration among these periods (ANOVA F = 20.388, df = 2, 140; p-value < 0.001). The precise
mechanism for the change is not revealed by the data set since lake ice duration integrates a number of
different ecological factors. Similar patterns of reduced ice duration correlated with increased local air
temperature have been noted for additional lakes over similar time scales (Assel and Robertson, 1995)
and shorter time scales (Wynne and Lillesand, 1993; Doran et al., 1996; Wynne et al., 1996; Anderson et
al., 1996). AVHRR satellite imagery has played a major role in determining ice cover duration dates for
lakes on the Laurentian Shield and satellite-determined dates were correlated well with available ground-
reference breakup dates (Wynne et al., 1996).
-------
o
400
375
350
325
300
275
250
225
200
175
150
125
100
75
50
25
0
y=1.1322x- 2030.9
R2 = 0.4259
Note: The moving average is based upon an 11 year computation.
1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Year (season ending year)
Figure 2.
1888 -1999 Houghton/Calumet seasonal snowfall amounts.
Seasonal snowfall at the Calumet-Houghton Ml airport (47°10'N
88°30'W). Both snowfall and variability increase over time. The
increased amount of snow over time may be an indication of climate
change (warming) over time. Snowfall data were obtained from the
Michigan State Climatology Program, Michigan State University and
the National Climatic Data Center CLIMVIS National Weather
Service Summary of the Day data.
Like the duration of ice
cover on a lake, the amount
of snowfall received at a site
is a phenological record that
has been maintained for long
periods in some locations.
The Calumet-Houghton area
of the Keweenaw Peninsula
on the south shore of Lake
Superior receives close to the
greatest seasonal amount of
snow of any location in the
continental North America.
All 1-year record of
snowfall in that area (Figure
2) reveals trends analogous to
those seen in the ice duration
data.
Three periods (1888-
1935, 1936 - 1974, and 1975
1999) apparently display
step increases in snowfall and
variability across time. There
is a significant difference in
snowfall among these periods (ANOVA F = 53.756, df = 2, 108; p-value < 0.001). Snowfall appeared
fairly constant across the first period (mean ± 95 % C. I. = 127.1 ± 7.2 in., C. V. = 0.196), increased over
the second period (mean ± 95 % C. I. = 186.1 ± 13.0 in., C. V. = 0.216), and is highly variable and
declining over the third period (mean ± 95 % C. I. = 225.328 ± 24.8 in., C. V. = 0.267). It is interesting
to note that Robertson (1989) examined the snowfall record for Madison WI from 1884 - 1988 and found
no trends. Madison is not located in the region of thermal influence of a large lake. Mid-continental
areas, not being subject to the interaction between seasonal changes in the heat budget of large lakes and
winter temperatures, may not reflect any differences in snowfall over time seen in the lee of Lake
Superior. As is the case with lake ice duration, the precise mechanism for a change in lake-effect
snowfall is not revealed by the data set since lake-effect snowfall also integrates a number of different
ecological factors.
The long-term trend in snowfall amounts in the Keweenaw Peninsula appears to parallel the duration
(number of days) of snow cover in central North America. Robinson (1991) noted a trend towards
increasing snow cover duration in the central United States Great Plains from the 1930's to the early
1970's. Robinson also noted that snow cover duration declined through the 1970's and 80's with the
lowest recorded snow cover occurring in the late 1980's. Changes in ice-cover duration (Robertson et al.,
1992; Assel and Robertson, 1995) suggest a cold period from 1855 ~ 1890, followed by an abrupt
warming around 1890 and a continued trend toward warmer temperatures. They report an additional
second wanning trend starting in the 1940's near Lake Michigan (Grand Traverse Bay) but not noticeable
near Lake Mendota (south central Wisconsin) until the mid-1970's.
Several authors have noted a large, positive trend in snowfall amounts to the lee of the Great Lakes
while no such trend is noted for areas farther inland or to the north and west of the lakes: (Namias, 1960;
-------
Eichenlaub, 1970; Eichenlaub, 1979; Brahamand Dugney, 1984; Harrington etal, 1987; Eichenlaub et
al., 1990; Leathers, 1993; Ellis and Leathers, 1996; and Leathers and Ellis, 1996). Other authors have
noted an earlier onset of springtime warming in the Great Lakes region based upon the analysis of long-
term records of water temperature (McCormick and Fahnenstiel, 1999), air temperature (Karl et al.,
1989; Boden et al., 1990; Hanson et al., 1992), lake ice cover (Hanson et al., 1992; Reycraft and Skinner,
1993; Assel and Robertson, 1995; Anderson etal., 1996) and onset of the active terrestrial vegetative
growth season (Keeling et al., 1996; Myneni et al., 1997). Therefore, the apparent increase in lake-effect
snowfall cannot be explained either by an increased duration of the winter season in the Great Lakes
region or by an overall increase in snowfall in the Great Lakes region.
Eichenlaub et al. (1990) attribute the increase in lake effect snow in the Upper Peninsula to the
colder than normal Januarys that occurred in the 1960's and 1970's. However, the increase in lake effect
snowfall started in the early 1930's and has occurred over a period when winter temperatures were
warmer than normal (1930 to the late 1950's) as well as colder. Clearly, the increase in lake effect
snowfall cannot be explained by changes in air temperature alone. Mean mid-summer temperatures in
Michigan are more persistent around the mean during the 1914-1987 period than mean winter
temperatures. For the period the 1930-1987, mean summer temperatures have displayed modulation (a
lessening of the diurnal temperature range) with a decrease in daytime highs and an increase in nighttime
lows. Eichenlaub et al. attribute this modulation to a decrease in the percent possible sunshine due to
increasing cloudiness in the summer.
Ellis and Leathers (1996) noted an upward trend in midseason lake-effect snowfall over the last
century. They used NCDC Summary of the Day daily snowfall data, hourly meteorological data, and
atmospheric data for the 32-year period from 1950-51 to 1981-82 and lake surface temperature and ice
cover data for the period 1966-88. Lake surface temperatures for two-week periods were used to
calculate lake surface to 850-mb lapse rates. They used an automated temporal synoptic index (TSI) to
identify weather situations conducive to lake-effect snowfall across New York and Pennsylvania. Their
principle components analysis within the TSI indicated the largest loadings (in decreasing order of
importance) were from terrestrial atmospheric variables: air temperature, dew point, pressure, wind
direction and speed and cloud cover. While they acknowledge the importance of surface water-to-
atmosphere transfer of heat (i.e.: seasonal reduction in lake-effect snowfall due to increasing ice cover on
the lake), and note the increase in lake-effect snowfall this century, they made no connection with the
possible increase in heat stored in the lakes.
A "heat budget" is a calculation of the amount of heat energy entering or leaving a lake. The most
commonly computed heat budget, the annual Birgean heat budget, is a measure of the heat energy input
from the lowest heat content in the winter to the greatest heat content in the summer. This heat budget is
expressed in units of cal cm"2 of lake surface area and actually is a measurement of the maximum heat
content gained by a lake in a year (Goldman and Home, 1983). A dimictic lake experiences two periods
of mixing each year. The lake is totally mixed and isothermal in the spring following ice-out and in the
fall prior to ice coverage. Between periods of mixing, the lake is thermally stratified (the lake has
different temperatures at different depths). The annual heat budget for a dimictic lake is calculated by
determining the difference between the minimum and maximum mean temperatures for the lake and
dividing that by the surface area of the lake. The morphometrically weighted mean temperature is
computed from the sum of the temperature of each depth stratum multiplied times the volume of each
depth stratum, divided by the whole lake volume. To accurately compute a Birgean heat budget, a series
of vertical temperature profiles must be obtained and accurate bathymetric data used to determine the
strata volumes (Wetzel, 1983; Lind, 1985). Often, heat budgets calculated for smaller lake systems with
relatively uniform bathymetry are based on a time series of vertical temperature profiles obtained at a
-------
single location, typically the maximum depth of the lake. Horizontal homogeneity of temperature is
assumed within each depth stratum. Assumptions of homogeneity may be warranted for small ponds and
single-basin kettle lakes but lakes with complex bathymetry and very large lakes present complex thermal
regimes that require more intensive sampling.
The mixing regime of a lake has been one method traditionally used to classify lakes. Hutchinson
(1957) finalized the most commonly used classification system based on mixing used to describe the
general patterns that usually occur in lakes. While the Hutchinsonian system works well for most lakes,
very large lakes still present problems in classification. Monomictic lakes have a thermal regime
characterized by continuous mixing throughout the winter season, except in years when the whole lake
surface freezes. Goldman and Home (1983) state "Typical monomictic lakes are the Great Lakes except
Lake Erie . . ." Kevern et al. (1998) state "When the Great Lakes don't freeze over, they commonly mix
throughout the winter and have only one extended mixing period from Fall to Spring and can be
classified as warm monomictic lakes". In contrast, Wetzel (1983) describes a warm monomictic lake as
having temperatures that "do not drop below 4° C", which excludes the Great Lakes. In the GLERL
publication ERL GLERL-81, Schneider, Assel and Croley (1993) state that the turnover behavior
exhibited by the Great Lakes "are a fundamental behavior of dimictic lakes" and Bennett (1978)
describes the seasonal cycle of vertical temperature in Lake Superior as "typical of dimictic lakes". The
Great Lakes often form a weak inverse thermal stratification in the winter and in some years their surface
entirely freezes, resulting in a strong inverse thermal stratification. The Great Lakes therefore should
properly be designated as dimictic in those years. Imboden and Wu'est (1995) find that lake mixing is
"too complex to be typified" in terms of their mixing regime and decline to use the terms "monomictic"
and "dimictic" at all.
"Turnover" describes a mixing condition when the entire lake mass is mixed and isothermal.
Regardless of the mixing classification applied to the Great Lakes and whether or not inverse
stratification has been established over the winter months, the surface temperatures of a large lake during
the spring and fall turnovers reflect the temperature of the entire body of water. Surface temperatures in
summer reflect only the temperature in the upper, mixed layer of water in the lake called the epilimnion.
Surface temperature does not reflect the temperature of the bottom, unmixed layer of water called the
hypolimnion. The hypolimnetic temperature of Lake Superior remains nearly constant in temperature
throughout the year (Millar, 1952; FWQA, 1968; Smith, 1972). The traditional assumption for the Great
Lakes is that variations in surface temperatures reflect differences in the depth as well as temperature of
the epilimnion. Figure 3 is a reprint of a figure (IJC, 1977) displaying the normal range of values for
lake-wide averaged surface and whole-lake mean temperatures of Lake Superior. The surface
temperature shows a much larger variation than the whole-lake mean since only about 15% of the entire
lake volume is a part of the mixed layer during summer stratification. Therefore, if the depth of the
mixed layer is known or can be assumed, measuring the change in summer surface temperatures should
provide an estimate of the yearly heat budget of the lake.
-------
0
§
CD
Q.
I
Range of
Lake - Average
Surface
Temperature
M
1 I r
A S O
N
D
Lake Superior: Seasonal cycles of lake-wide average surface temperature (triangles)
and volume mean temperature (circles), based on 19 surveys in the
April-November period (CCIW files) and 4 surveys in December-January
(7). Each point is a cruise mean value plotted without regard to year of
observation. The curves labelled "range" define the approximate limits
within with the lake-wide average surface temperature can occur.
Figure 3. Lake-wide average surface and volume mean temperature in Lake Superior.
Reprint of a figure from the Physical Limnology section of the International Joint
Commission 1977 book "The Waters of Lake Huron and Lake Superior".
-------
Bennett (1978) used a
figure (reproduced here as
Figure 4) to illustrate the
seasonal cycle of the lake-
wide averaged vertical
distribution of temperature in
Lake Superior. The lake-wide
averaged vertical temperature
distribution shows a
thermocline forming at around
10 m in depth in mid-summer.
Bennett shows the
thermocline remaining at this
depth while the epilimnetic
temperatures increase until
late in the stratified season
when the mixed layer begins
to deepen as epilimnetic
temperature fall.
£ 100
f
0)
Q
150
200
250
Figure 4.
Seasonal cycle of the lake-wide averaged vertical distribution of
temperature (°C) in Lake Superior. The semi-annual alternation
between periods of stratification and of extensive vertical mixing is
typical of dimictic lakes. From E.B. Bennett (1978).
In very large lakes, the summer depth of the warm, isothermal mixed layer (the epilimnion) is based
upon wind mixing with an upper limit set by latitude and the Coriolis effect. Gorham and Boyce (1989)
demonstrated mathematically that the thickness of the mixed layer (Ed) of a lake increases with
increasing lake size until limited by the earth's rotation (Coriolis effect). In large lake systems, effects
such as water transparency and short-term variations in solar insolation are relatively less important than
in small lakes in determining the heat budget of the lake. Fee et al. (1996) studied a large number of
Canadian Shield lakes of various sizes over a 23-year period and confirmed this prediction of an absolute
upper limit for Ed. Fee et al. found that the largest lake in their study (Nipigon, 484,800 ha surface area)
had an Ed no larger than Trout Lake (34,700 ha). Several studies (Kling, 1988; Mazumder et al, 1990;
Mazumder and Taylor, 1994; Schindler et al., 1996) have stressed the importance of water clarity in
determining Ed. Sterner (1990) showed that in lakes > 1250 ha, Ed increases solely as a function of the
area of the lake. Fee et al. believe that in large lakes (> 1250 ha), physical forces of wind stress and the
earth's rotation determines Ed independently from water transparency and in "very large lakes", the
Coriolis effect alone determines the mean Ed. Fee et al. don't give their estimate for the size that qualifies
as "very large" but Boyce (1974) defined as series of internal motions in lakes that are only found in
lakes > 5,000 ha (length scales > 10 km). These large-scale processes such as internal waves, coastal
currents, and upwelling-downwelling make individual in situ measurements of Ed in very large lakes
highly variable (Mortimer, 1993). However, from a heat budget perspective, the whole-lake averaged
mean depth of the mixed layer over the course of a spring-fall cycle may be stable enough to be
considered a constant for lakes large enough that the earth's rotation determines Ed.
Figure 5 reproduces a figure from Schertzer and Sawchuk (1990) that displays daily vertical
temperature profiles for Lake Erie and Lake Ontario for 1983. The temperature values were interpolated
between sample stations with a station density averaging 1:400 km2. At this spatial and temporal
resolution, the thermocline is remarkably stable, particularly for the deeper Lake Ontario.
In 1983, an unusually warm year, Schertzer and Sawchuk found that Lake Erie and Lake Ontario
displayed higher than normal surface temperatures with a thermocline that formed earlier and deepened
later. In the summer, the thermocline depth fell within expected values calculated from long-term (1967-
-------
1982) data and was found at a shallower depth than usual in the fall. The fall thermocline is shallower
because the deepening of the thermocline prior to the breakdown of stratification and vertical mixing of
the water column occurred later than usual.
a. Lake Erie
22.2 24.3 24.3 23.6
23.0
8.0
b. Lake Ontario
0
23.5
16.6
5.2
15.9
Figure 5.
Thermal structure of the lower Great Lakes. Computed daily vertical temperature profiles
(thin lines) for the central basin of (a) Lake Erie and (b) Lake Ontario, 1983. Heavy vertical
lines represent temperature profiles evaluated from Lam and Schertzer's (1987) thermocline
model for Lake Erie and Simons's (1980) thermocline model for Lake Ontario. Solid circles
represent observed values determined from lakewide ship cruises. Calculated depths of the
base of the epilimnion and top of hypolimnion are illustrated with dark horizontal lines.
Numerical values represent observed surface and bottom temperatures. (From Schertzer and
Sawchuk, 1990.)
DeStasio et al., (1996) used computer simulations of four study lakes based on four general
circulation model (GCM) scenarios. The lakes ranged in surface area from 36.7 - 3,940 ha and in
maximum depth from 19 - 35.7 m. Summer stratification timing arid intensity changed in all lakes but
stratification depths did not change much in the larger lakes. Hypolimnetic temperatures were variable
among lakes and scenarios. Their computer modeling suggested that changes in water temperature due to
climate change would be mostly confined to epilimnetic waters.
Robertson and Ragotzkie (1990) used both dynamic and statistical approaches to model the thermal
response of lakes to increased air temperatures due to climate change. They found that increases in air
temperature resulted in increases in epilimnetic temperature and the length of the stratified season but
-------
very little change in hypolimnetic temperature or thermocline depth in large lakes. The thermocline
depth was shallower during late summer but this was due to an increased period of stratification rather
than an absolute change in thermocline depth. The stability of the water column increased due to the
larger response in epilimnion temperatures than in hypolimnetic temperatures.
In contrast, McCormick (1990) predicts that increased air temperatures will reduce vertical mixing
and further increase epilimnetic warming due to a smaller mixed volume. Figure 6 reproduces a set of
figures from McCormick (1990) that shows the seasonal cycle of the lake-wide averaged vertical
distribution of temperature for Lake Michigan along with a series of distributions derived from models
depicting expected changes due to global warming. While McCormick based his computed changes in
epilimnetic temperature upon changes in epilimnetic depth, his figures display little change in
thermocline depth.
150
JFMAMJJASOND J - F ' M ' A " M ' J ' J ' A - S " O ' N ' D
Time (months)
Figure 6. Simulated temperature fields (low-pass filtered) under present (Base) and possible
future conditions (determined by the global climate models, GISS, GFDL, and OSU).
Contour interval is 1 °C for temperatures below 6°C and 2°C for warmer temperatures; the 10°C
isotherm is dashed. (From McCormick, 1990.)
The seasonal difference in thermocline pattern expected with global warming that is predicted by the
authors previously discussed is illustrated in Figure 7.
10
-------
100
Julian Date
125 150 175 200 225 250
275
300
325
350
u •
5
•? 10
3 15
t20
25
o 30
a 35
40
Ac
vv
" " ""
Current Base of Epilimnion Depth
Post-Warming Base of Epilimnion Depth
Figure 7. Hypothetical 10°C isotherm depth with climate warming.
"T^
Hypothetical relationship betwee
the thermocline depth of a large lake before and after climatic warming of that lake. Under this
hypothesis, the depth of the mixed layer remains constant while the period of time the lake is
stratified increases.
Few large-scale studies of the vertical temperature structure of Lake Superior have been done. Most
reports (e.g. Smith and Ragotzkie 1970; Smith, 1972; Hoopes et al., 1973) are based on a series of
bathythermograph temperature cross-sections obtained from cruises and in situ temperature recorders
obtained in the 1960's during the Federal Water Quality Administration (FWQA) study. Bennett (1978)
summarized 19 separate studies by the Great Lakes Institute and Canada Centre for Inland Water over the
period 1964-73. Millar (1952), the FWQA study, Sloss and Saylor (1976), and Bennett all found similar
patterns in the vertical thermal structure of Lake Superior. In general, an initial mid-lake epilimnion
depth of about 10 m is established by mid-July. The epilimnetic depth remains essentially the same
during mid-summer stratification and begins to increase in mid-September through the fall. Lake
turnover and isothermal temperatures typically occur in late November or early December. However, the
thermal profiles used to make these generalizations are inadequate. Sloss and Saylor (1976) characterize
usefulness of the FWQA temperature data as "limited by many mechanical failures of the recording
devices and uncertainties in the timing of many records". Other studies are limited in geographic and
temporal scope. Further research is needed to determine the seasonal variations of vertical temperature
structure in Lake Superior as well as other large lakes and how those variations relate to the surface
temperature of those lakes.
Large lakes, with length scales larger than 10 km, are subject to surface and internal seiches, coastal
currents, upwelling and downwelling, and wind-driven horizontal waves (Boyce, 1974). Surface
temperatures of very large lakes (length scales larger than 100 km) exhibit a high degree of horizontal
heterogeneity during the summer (see Appendix, Figure Al for a mid-summer thermal AVHRR image of
Lake Superior). The size and complexity of large lakes means that obtaining the data needed to calculate
an averaged value for the surface temperature of the lake is a major and costly undertaking using
conventional (shipboard and deployed) instrumentation. Remotely sensed thermal data may be the only
economical means of estimating whole-lake surface temperatures for many large lakes in North America.
One of the goals of this research is to try to quantify the relationship between surface and whole lake
temperatures in large lakes. The research proposed in this research plan seeks to address the problem of
relating surface temperatures to thermal structure at depth by using existing surface and vertical
temperature data from large lakes where long-term datasets exist and comparing those values to remotely
sensed values from CoastWatch AVHRR imagery. An existing dataset from a large lake in the Great
Lakes watershed is used below to illustrate the measurements to be used.
11
-------
I obtained vertical temperature measurements at 7 -14 day intervals during the 1995-1997 summer
seasons at Lake Michigamme (lat./long. 46.5° N, 88.1 ° W, elevation 474 m). Lake Michigamme is a
17.6 km2 multi-basin lake in the north central Upper Peninsula of Michigan, USA, located in the upper
watershed of Lake Michigan. The surface topography of Lake Michigamme is dominated by a long,
narrow, main basin with a maximum depth of 22 m and an east-west fetch of 10.73 km that lies in a
valley having an east-west orientation. The prevailing wind direction is from the west and the lake is
noted for its vigorous surface
waves. The Ojibwa Native 30
American name from which
the anglicized word 25
"Michigamme" is derived ^*
means "streaked water1' and t_ 20
the foam lines characteristic &
of Langmuir circulation are fg 15
evident on Lake Michigamme g>
with only light winds. The E 10
central basin epilimnion is H
characterized by vigorous 5
mixing. The vertical
temperature profiles obtained o
in the central portion of the
main basin are representative
of temperature patterns seen
in highly mixed waters. The Figure 8.
1995 seasonal mean
temperatures of the
epilimnion and hypolimnion
of Lake Michigamme are
illustrated at right (Figure 8).
Temperature change was
mainly confined to the
epilimnion. This pattern was
repeated in 1996 and 1997.
Mean Epilimnion Temperature (°C)
- - Mean Hypolimnion Temperature (°C)
150 160 170 180 190 200 210 220 230 240 250 260 270 280 290
Julian Date
Summer 1995 Lake Michigamme temperature. Mean epilimnetic
and hypolimnetic temperatures measured by a series of vertical
temperature profiles from the central basin of Lake Michigamme,
Upper Peninsula, Michigan. From Julian Date 158 (June 7) through
255 (September 12) most of the change in temperature in the lake
occurred in the epilimnion. The dip in the early summer epilimnetic
temperature is due to change in the depth of the mixed layer from
June 23 to July 11 when the depth of the epilimnion increased from
2 to 5 m. The mean temperature of the epilimnion varied over the
course of the summer while the temperature of the epilimnion
remained nearly constant.
While surface temperatures in Lake Michigamme are not well correlated to the overall mean
temperature of the whole water column at any point in time (R2 = 0.1141, Figure 9), surface temperature
was highly correlated to the bulk epilimnetic temperature (R2 = 0.9737, Figure 10).
Figures 11 and 12 plot the seasonal curves of the surface and mean temperature of a 12 m deep
column of water in the central basin of Lake Michigamme for the years 1995-1997. The surface
temperatures were more variable than the water column temperature for 1995-1997. The water column
was warmer in 1995 than 1996 and 1997.
To look at a single seasonal value for each temperature parameter measured, the water column mean
and surface temperatures were integrated over the season and the seasonal mean computed. The values
obtained for each year were plotted (Figure 13).
12
-------
30
25
20
15
10
5
0
Figure 9.
Q. o>
o £
«!
"5 0
£ Q.
y = 0.2073x-M3.112
R2 = (
10 15 20
Surface Temperature (°C)
25
30
1995 -1997 Lake Michigamme surface vs. whole depth mean
temperatures. Correlation between surface temperature and the mean
temperature of the entire water column in Lake Michigamme for the years
1995-1997.
30
p 25
O !—' on
as a) 20
II
iu
15
10
5
y = 0.9394X + 0.8772
R2 = 0.9737
10 15 20
Surface Temperature (°C)
25
30
Figure 10. 1995 -1997 Lake Michigamme surface vs. epilimnetic temperatures.
There was a very tight correlation between the surface temperature and the
bulk mean epilimnetic temperature of Lake Michigamme during the summer for
the years 1995-1997.
25
t? 20
a
¥ 15
ra
fc 10
Q.
Whole Lake Mean Temp
-fr-1995
-•-1996
1997
140
160
180
200 220 240
Julian Date
260
280
300
Figure 11. 1995 -1997 Central Basin Lake Michigamme mean temperature of the
whole water column. Seasonal curves showing the changes in the mean
temperature of a 12 m column of water in the central basin of Lake
Michigamme during the summer seasons of 1995-1997.
13
-------
30
O 25
20
8. 10
I 5
0
Surface Temperature
-*-1995
140 160 180 200 220 240
Julian Date
260
280
300
Figure 12. 1995 -1997 Central Basin Lake Michigamme temperature of the water
surface. Seasonal curves showing the changes in the surface temperature
in the central basin of Lake Michigamme during the summer seasons of
1995-1997.
o 20-5
20.0
2 Q- 1Q c;
ro E
«J <5
T g 19.0
18.5
1995
1997
1996
16.0 16.5 17.0 17.5 18.0
Time-Integrated Whole Lake Mean Temperature (°C)
18.5
Figure 13. Summer 1995 -1997 Lake Michigamme temperatures. Comparison of the
seasonal integrated values for the surface temperature and mean
temperature of the water column in the central basin of Lake Michigamme
during the summer seasons of 1995-1997.
While there are only three years in this comparison, it appears that a seasonal integrated mean value
for the surface temperature provides a reasonable predictor for the seasonal mean temperature of the
column of water below the surface point measured. The surface temperature, integrated over the course
of an entire season, appears to provide an indicator of the temperature of the entire body of water.
Therefore, it appears reasonable that annual changes in heat content of large lakes may be detectable
from measurements of surface temperatures alone. Large, well-mixed lakes may therefore be good
analogs for climate change if a long-term record of seasonal surface temperature values can be compared
over time.
Proposed Research
The proposed research will use a series of NOAA Advanced Very High Resolution Radiometer
(AVHRR) thermal images of Lake Superior to construct a seasonal estimator of the summer mean
surface water temperature for 1994. This indicator value will be compared to the value generated by the
14
-------
NOAA GLERL GLSEA automated mapping of surface water temperatures (Schwab et al., 1999). Either
of the methodologies used in producing the GLSEA product or the proposed estimator could be applied
to smaller lakes that are not currently included in the GLSEA dataset or to the Great Lakes for periods
not currently included in the GLSEA dataset. Both the GLSEA and the proposed indicator are based
upon AVHRR imagery.
The AVHRR data set has a high temporal resolution. Visible band imagery is obtained once daily
and thermal imagery is obtained during both daytime and night passes from each operational NOAA
satellite. The satellites have a total field of view that allows for images of the entire Lake Superior area
to be collected in a portion of a single image. The high temporal resolution allows for events that occur
over time to be tracked closely with multiple images. The impact of weather interference from cloud
cover is reduced (but not eliminated) by the opportunity to obtain images every day. Coastal fog and
cloud cover varies seasonally over Lake Superior. Weather interference over Lake Superior is less
common during the spring and summer months.
The main disadvantage to the AVHRR imagery is the lower radiometric resolution and spatial
resolution of the imagery. For NOAA 6 -14, only two bands of reflected light, Band 1 (red, 0.58 - 0.68
yum bandwidth) and Band 2 (near-IR, 0.72 -1.10 /^m bandwidth), are available. Band 3 (3.55 - 3.93 /u.m
bandwidth), Band 4 (10.50 -11.50 /^m bandwidth), and Band 5 (11.50 -12.50 //m bandwidth on odd
numbered NOAA satellites and a Band 4 repeat on even numbered NOAA satellites) are thermal
wavelengths of electromagnetic radiation (emitted energy). The pixel size at nadir is 1.1 X 1.1 km at a
nominal altitude of 833 km (Kidwell, 1991). This coarse spatial resolution means that only large-scale
differences can be mapped and that the values recorded for a pixel are an averaged value over a large
area. This loss of spatial resolution is the price that is paid to obtain high temporal resolution. At the
whole-lake scale of study proposed in this research, the integration of small surface irregularities due to
pixel size acts as a valuable smoothing function that has no negative effect on the total or mean
temperature sensed.
The images will be acquired from the CoastWateh Region AVHRR database, using the NOAA
Coastal Active Access System (NCAAS). The CoastWateh high-resolution images of Lake Superior are
mapped to a Mercator projection at a nominal resolution of 1.3 km. The images are free and will be
obtained via INTERNET using a binary file transfer protocol (ftp). A series of separate compressed
image files will be obtained for each image date. Bands 1 and 2 are separately available as *.scl and
*.sc2 images respectively and Bands 3, 4, and 5 are combined to create single SST (sea surface
temperature) images. The SST images are thermal images where a multi-channel sea surface temperature
(MCSST) or a nonlinear sea surface temperature (NLSST) algorithm has been applied to produce the
image. Bands 3, 4, and 5 are separately available for nighttime acquisitions but no Band 1 and 2
(reflected light) data is simultaneously available for those acquisitions (Pytlowany, 1994). The MCSST
and NLSST algorithms used to create the thermal CoastWateh images have changed over time since
1990. Different algorithms have been used with different satellites over the time covered by
CoastWateh. For the Great Lakes CoastWateh images, a linear split window algorithm is used for the
MCSST image and a nonlinear split window algorithm is used for the NLSST image. Both daytime and
nighttime cloud masks are currently available for CoastWateh imagery, but the periods that cloud masks
are available vary among CoastWateh Regions. To maintain continuity in the data set over time and
between regions and to ensure that reflected light imagery would be available to visually compare with
the thermal imagery for cloud masking, only daytime acquisitions will be used.
The time-series of thermal images of Lake Superior will be georectified to the same base map and
each image will be thermally registered to the same scale (see Appendix, Figure Al for an example
15
-------
image). This will allow each image to be directly compared with all other images from the same
sequence and for thermal changes over time to be clearly tracked. The mean of all pixel thermal values
on an image date provides an estimate of the mean surface water temperature of the entire lake on that
date. The sum of all of the temperature values over a given period divided by the number of days in the
period provides a mean estimator of the surface water temperature for that period. This thermal value is
an analog estimator for the whole lake temperature because it can be expected to vary along with the total
epilimnetic heat but cannot be used to directly calculate the total epilimnetic heat without additional
information.
The thermal value for each lake pixel provides an estimate of the averaged bulk epilimnetic water
temperature beneath the surface area of the pixel remotely sensed. To determine the actual total heat
content held in the epilimnion would require knowledge of the depth of the epilimnion. This knowledge
is not obtainable via remote sensing. Vertical temperature profiles of lakes are typically available from
other sources from only a few locations (usually the deepest depth of the main basin) on a few lakes.
However, since the surface total temperature will vary over time.along with the total epilimnetic
temperature, the surface temperature of very large lakes provides an estimator of the total epilimnetic
heat content if the depth of the epilimnion is presumed to be a seasonal constant for that lake over time.
Integrating the surface temperature values over a season will allow for an estimate of the seasonal heat
content gained and lost for the entire lake. Comparing the integrated thermal values calculated for a lake
over time will allow for assessment of possible changes in the thermal budget of that lake over time.
An example of an analog measurement commonly used in limnology is the secchi disk depth. The
secchi disk depth is a complex interaction variable based upon the surface solar irradiance, atmospheric
conditions, DOC, particulate
concentration and particle size.
Knowing the secchi disk depth
on a given date on a given lake
yields no information about
the nature of the physical and
chemical conditions of that
lake. However, a series of
secchi disk measurements over
a long time at a single lake can
yield valuable clues about
changes in those conditions
because the secchi disk depth
is correlated with physical and
chemical conditions (e.g.
Goldman and Home, 1983;
Jassby etal, 1999; Figure 14).
15
20
25
30
35
40
68 70 72 74 76 78 80 82 84 86 88 90
Figure 14.
Lake Superior will be used
as the initial study lake for
several reasons. First, Lake
Superior has the largest
surface area of any freshwater
Annual average secchi depth (± 1 s.d.) in Lake Tahoe. Cultural
eutrophication in ultraoligotrophic Lake Tahoe. Long-term change in
water transparency in Lake Tahoe. The secchi depth has
decreased at an average rate of about 0.5 m annually. Three-
quarters of the variation (regression coefficient r2 = 0.75) in secchi
depth over this time is explained by the change in time. Algal
growth in the lake has increased steadily over this time. Modified
from Goldman (1993). (From Home and Goldman (Limnolgy
1994).)
lake in the world. Wide field-
of-view remote sensing provides a clear advantage to traditional sampling methods when studying large
areas. Second, there are three seasonally moored buoys in mid-lake positions in Lake Superior
16
-------
(Figure 15). In addition, there are four fixed Coastal-Marine
Automated Network (C-MAN) stations along the Lake
Superior shoreline that record meteorological data. The
moored buoys record a number of parameters
including the water temperature at 0.6 m or
1 m depths on a near-continual basis
during the ice-free navigational
season (Figure 16).
The moored buoys
provide a long-term "ground"-
truth data set with which
remotely sensed temperatures
can be compared. The
moored and C-MAN buoy
data are available from the
U.S. National Oceanographic
Data Center National Data
Buoy Center
(NOAA/NODC/NDBC). The
moored buoy 45001 (central
basin, Lat. 48.x N. /Long.
87.x W.) has been recording
temperature data since 1979.
Buoy 45004 (eastern basin,
Lat. 47.x N. /Long. 86.x W.)
has been recording
temperature data since 1980.
Buoy 45006 (western basin,
Lat. 47.x N. /Long. 90.x W.)
has been recording
temperature data since 1981.
The exact positions of these
buoys and their times-in-place
vary slightly from year-to-
year (Figure 15 shows the
1994 locations).
Calumet-Houqhton Airport
Figure 15. Lake Superior mid-lake moored buoy locations. 1994NOAA
moored buoy locations for 45001 (48°03'N 87°46'W), 45004
(48'03'N 87°46'W), 45006 (48°03'N 87°46'W), and Calumet-
Houghton airport (47°10'N 88°30'W).
p
0)
•*-•
2
0)
Q.
E
0)
200
205
210
Julian Date
215
220
Lake Superior is one
of the lakes included in the
GLSEA dataset. Surface
temperature values
determined during this
research can be compared to
analogous values computed
Figure 16. Lake Superior NDBC buoy 1994 water temp. Short time cycle
variability of Lake Superior buoy sensed temperature. A 20-day
period in mid-summer 1994 is shown. Between buoy temperature
differences reflect differences in the heating regimes at the buoy
locations. Buoy temperatures show variations on a diurnal and
several daytime cycles as well as on a seasonal basis. Remotely
sensed images over a season will integrate these temporal as well
as spatial variations.
for the same period using the
same base imagery but employing a different technique. This "air-truth" will complement the "ground-
truth" from the buoys.
17
-------
The final reason for using Lake Superior as a target of the initial research is that the lake-effect snow
impact downwind from Lake Superior is well recognized. The lake-effect from Lake Superior leads to
substantial local accumulations of snow in regions of the Upper Peninsula of Michigan that are far in
excess of snow levels seen farther inland. Annual and monthly snowfall records exist for lake-effect
snow areas that cover the periods of buoy and AVHRR data and before. These data are available through
the U.S. National Weather Service Midwestern Climate Center, the Michigan State Climatologist, and
the NOAA National Climatic Data Center. The seasonal amounts of snow received at recording stations
will be compared to the epilimnetic temperatures recorded at the moored and C-MAN buoys and the
thermal image estimator for epilimnetic temperature. Air temperatures measured at the meteorological
recording stations will be compared to the snowfall amounts at a series of different temporal scales to try
to shed more light on the complex interaction of environmental factors that influence the amount of lake-
effect snowfall (Figure 18).
Robertson (1989) found that since 1940, the year following the onset of an El Nino was directly
associated with unusually short ice duration on Lake Mendota and unusually warm spring air
temperatures. Anderson et al. (1996) studied lake ice breakup dates for 20 Wisconsin lakes from 1968 -
1988 and found a recent warming trend (toward earlier breakup dates). They also found a strong
influence of El Nino events on ice breakup, particularly in the southern area of their study but found no
significant relationship between total winter snowfall and the year in the El Nino-Southern Oscillation
(ENSO) cycle in the non-lake-effect areas studied. Assel (1998) and Assel et al. (1985) noted the
relationship between the ENSO cycle and percent ice cover on the Great Lakes. Mild winter
temperatures associated with the ENSO cycle and below-average maximum ice cover are correlated. The
difference between the maximum ice cover percentage between strong ENSO event years and the base
period average was statistically significant. The above results suggest there is a regional-scale climatic
response to El Nino events that is evident when viewing lakes and the regions they thermally impact.
The proposed research will look to see if there is a correlation between the thermal value computed for a
lake and the ENSO cycle.
A series of *.scl, *.sc2, and *.sdl CoastWatch images have been obtained for 36 dates from April
through October 1994. These images have been processed per the methods listed in the Technical
Approach below and will be used as the initial image set for the project. The MCSST temperatures
remotely sensed in these images at the locations of the NDBC moored buoys have been compared to the
buoy temperatures sensed on those image dates. The mean temperatures of the nine-pixel block centered
on the buoy location are highly correlated with the buoy temperatures (Figure 19; R2 = 0.974, n = 99
image locations with buoy data and image locations felt to be non-cloud impacted).
The majority (83.8 %, 83 out of 99) of the differences between buoy temperature and pixel
temperature are less than ± 1.0 ° C with 97.0 % (96 out of 99) of the differences less than ±1.75° C.
There is a negative temperature skew to the remotely sensed temperature data, with 15 of the 16
differences greater than 1.0 ° C being negative (Figure 20). The mean difference (buoy-MCSST) was
-0.36 (S.D. = 0.53). The values obtained through the MCSST algorithm and the cloud masking
techniques compare favorably with those from other researchers. Richards and Massey (1966) found that
micro-surface "skin" temperatures recorded with an infrared thermometer mounted on a research vessel
averaged 0.4° C less than the simultaneous temperature recorded at the 0.15-0.30 m depth. Richards et
al. (1969) reported the "skin" minus "bulk" temperature difference as - 0.6° C when comparing airborne
sensors to surface water. Schwab et al. (1992) reported satellite-derived temperatures that were
consistently 1.0-1.5° C below the moored buoy temperatures for the CoastWatch SST imagery of the
Great Lakes from 1990 and 1991. Pichel et al. (1994) reported a difference of - 0.8° C for the
CoastWatch SST imagery compared to fixed buoy measurements at cloud-free ocean sites.
18
-------
360
315
270
I
% 225
o
CO
180
135 -
90
La Nina Winters
For Entire Dataset:
y = 8.7419x+153.32
R2 = 0.0216
El Nino Winters
For El Nino Winters:
y= 11.587x + 83.681
R2 = 0.566
100 120 140 160 180 200 220 240 260 280 300
Julian Date
Figure 17. 1989 -1998 Lake Superior buoy 45001 water temperature.
Seasonal bulk water temperatures at Buoy 45001 from 1989 -1998.
Note how the May through mid-July temperature profiles are very
similar for all years except the unusually warm year of 1998.
Typically, the yearly differences in heat content of the bulk
epilimnion of the lake at Buoy 45001 are reflected in differences in
water temperature during the mid- to late summer and early fall.
0 1
4 5 6 7 8 9 10 11 12 13 14 15
Water Temperature (°C)
Figure 18. 1987-88 to 1998-99 seasonal snowfall vs. prior October water
temperature at buoy 45001. Preliminary analysis of the
relationship between the late-October Lake Superior bulk water
temperature measured at Buoy 45001 and the amount of snowfall
received at Calumet-Houghton Ml over the following winter. The
small size of the data set makes statistical interpretation risky but
there appears to be no direct relationship between the warmth of
the lake in the fall and the amount of snowfall. However, there
does appear to be a complex interaction dependent upon the
severity of the winter. It appears that the interaction between the
ENSO/La Nina cycle, seasonal air temperatures, and the heat
content of the lake may be a major factor for snowfall generated.
Comparisons over time of fall water temperatures, winter air
temperatures, and lake-effect snowfall amounts are needed to
study this interaction.
-------
o
O
i"
+*
2
0>
Q.
E
0)
"3
x
E
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
ro
o
y = 1.0015x + 0.3763
* = 0.974
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Buoy Temperature (°C)
o
£
3
0)
Q.
£
3.5-
3.0.
2.5
2.0 :
1.5
1 0
0.5
0 0
-0.5
-1 0
-1.5.
-2.0^
-2.5.
-3.0 •
-3.5
-d. n ,
•
i ,: : i ,,;;;.•.
» ":: . • • ""• **|" * * ' \ ' :
"* t • » •*
• •
Figure 19. 1994 AVHRR CoastWatch MCSST pixel vs. buoy temperature.
(n = 99 comparisons between 35 summer images matched
with Buoys 45001, 45004 & 45006.) Correlation between the
mean MCSST temperature sensed at a 9-pixel array centered at
the moored buoy locations 45001, 45004 and 45006 and the
recorded buoy temperature recorded closest to the satellite image
acquisition time. Values were included for those dates where
buoy data existed and visual examination of the *.sc1 , *.sc2 and
*.sd1 CoastWatch images indicated no cloud interference (n=99).
120 140 160 180 200 220 240 260 280
1994 Julian Date
Figure 20. 1994 Lake Superior buoy - pixel temperatures, (n = 99
comparisons between 35 summer images matched with
Buoys 45001, 45004 & 45006.) Magnitude and direction of the
Buoy - Pixel temperature error over time. The value displayed is
the difference between the mean MCSST temperature sensed at a
9-pixel array centered at the moored buoy locations 45001, 45004
and 45006 and the recorded buoy temperature recorded closest to
the satellite image acquisition time. Values were included for
those dates where buoy data existed and visual examination of the
*.sc1, *.sc2 and *.sd1 CoastWatch images indicated no cloud
interference (n = 99).
-------
Clouds and other forms of atmospheric interference are the greatest single source of error in the
interpretation of thermal imagery. Clouds have the effect of either blocking the thermal image entirely or
reducing the temperature sensed at a pixel if the area is covered with thin clouds. The effect of clouds
will have to be eliminated if whole-lake thermal values are to be calculated. This can be accomplished
by selecting only cloud-free images. However, for an area as large and as cloud-impacted as Lake
Superior, this would result in a calculation based on only a few images per season. This research
proposes to process the georectified time series of images to remove the thermal effect of clouds. The
thermal value for each cloud-impacted pixel will be estimated with an algorithm derived by comparing to
the non-cloud thermal values of the same pixels sensed temporally before and after the cloud-impacted
pixel date. This will allow for the thermal values of cloud-influenced pixels to be estimated by
interpolation. Adjacent lake areas that are not cloud-impacted will be simultaneously sampled to scale
the thermal interpolation. Cloud-masking algorithms have been developed for AVHRR imagery and
CoastWatch Level 3 product images include cloud masks for years after 1994 (Cayula and Cornillon,
1996; Maturi and Pichel, 1993 and 1994; Vazquez, Perry and Kilpatrick, 1998). Most current methods
used to replace clouded pixels in NOAA-AVHRR images work best for scenes, which are almost cloud-
free and are based on horizontal techniques used on a single image (Addink and Stein, 1999). The use of
cloud-masking algorithms will be explored as a part of the initial phase of the research project.
The proposed thermal interpolation technique following cloud masking will allow for the thermal
removal of clouds for images where cloud cover is not too extensive (less than 40 %) and georectification
of the images is accomplished. The goal is to obtain at least one image for every 7-10 day period over
the course of the season. Due to the commencement of lake-effect cloud formation and precipitation in
the fall, thermal remote sensing sampling seasons will typically end in October or November. To the
greatest extent possible, images will be obtained to cover the same time that buoys are in place. The
buoys are typically in place from May through October each year. The entire season (mean) temperature
or only the late-fall temperature may be more important to the following winter's lake-effect snow
amount. Values for both parameters will be calculated from the data.
The "thermal indicator season" as defined for this research will start on the latest date that moored
buoys were deployed and end on the earliest buoy removal date during the period 1991-1999. This will
allow for a uniform time series of values to be compared between years. The ideal situation is where the
thermal sampling season for each year is bracketed by cloud-free imagery, isothermal lake conditions,
and moored buoys in place. These conditions are likely to be met in the spring but are likely to be
problematic in the fall due to the removal of buoys prior to isothermal lake conditions and cloud-
interference. The seasonal ending date for among-year comparisons is likely to be a compromise value.
It should be noted that including the estimated pixel thermal values in the mean thermal value
computed for the whole lake on a cloud-impacted date reduces the confidence level in the thermal value
computed for that date. However, including the estimated pixel thermal values in the mean thermal value
should also make the mean value closer to the real value of the lake on that date. The ability to use
images impacted by clouds increases the number of images that can be used when calculating the
seasonal temperature changes in the lake. Including the estimated pixel values for cloud-impacted dates
therefore should increase the accuracy of the yearly integrated thermal value computed.
The seasonal thermal values calculated with the proposed technique will be compared to average
surface temperature values calculated from the NOAA Great Lakes Environmental Research Laboratory
Great Lakes Surface Environmental Analysis (GLSEA) for corresponding periods. The GLSEA is a
digital map of the Great Lakes surface water temperature and ice cover. This map is produced daily at
the NOAA Great Lakes Environmental Research Laboratory (GLERL) in Ann Arbor, Michigan through
21
-------
the NOAA CoastWatch program (Schwab et al., 1999). The lake surface temperatures are derived from
NOAA polar-orbiting satellite imagery obtained through the Great Lakes CoastWatch program. Lake
surface temperatures are updated daily with information from the cloud-free portions of the previous
day's satellite imagery. If no imagery is available, a smoothing algorithm is applied to the previous day's
map. Currently, the only full calendar years for which GLSEA data are available for Lake Superior are
1995 -1998. A preliminary comparison of the sum of the GLSEA average surface water temperatures
for the calendar years of 1995 - 1998 and the trapezoidal integration of the surface water temperatures
sensed at NODC Buoy 45001 for Julian dates 125-290 of the same years shows these values to be highly
correlated (linear regression y = 0.8527x - 642.59, R2 = 0.9756). While this data set is too small to draw
sweeping conclusions, the strong correlation supports the hypothesis that most of the annual surface
temperature difference among years can be expressed by differences in surface temperatures during the
periods within Julian dates 125-290.
Time series analyses will be performed on the GLSEA dataset and the snowfall records for
meteorological stations located both in the lee of Lake Superior and outside of areas affected by lake-
effect snowfall. Cross-correlation will be used to compare snowfall, Southern Oscillation Index (SOI)
values, air temperature, and water temperature during the 1992-1999 study period. Seasonal behaviors
will be studied first and autoregression will be used to check for dependency over time. Categorical
analysis (CART decision tree) will be used as a "nonparametric multivariate" approach to relate the
differences in observed snowfall to temperature and other differences by season. Parametric multivariate
analysis may also be appropriate if the necessary assumptions required are met in the dataset. Statistical
analyses will be carried out with SAS and SYSTAT software in cooperation with Maliha Nash, an EPA
BSD statistician located at the LEB-LV lab.
Detailed surface and/or vertical temperature data are lacking for most lakes large enough to have
surface temperatures monitored by AVHRR imagery. There are a huge number of lakes > 1,000 ha in the
North American mid- to high latitude region where the effects of climate change due to greenhouse
warming are expected to be most apparent. These lakes form the potential pool of research targets using
the techniques to be developed with this proposal. Studies of the apparent changes in thermal
characteristics of these lakes can be compared with previous and current studies looking at related
characteristics of the same lakes. An archive of 1980 to current AVHRR and GOES images has already
been assembled to study the relationship between lake ice and climate in mid- to high latitudes in central
North America (Wynne and Lillesand, 1993; Wynne et al., 1995; Wynne et al., 1996; Wynne et al.,
1998). Collaborative relationships with other researchers and agencies will be pursued as a part of the
proposed research.
Records of surface and vertical temperatures will be pursued and obtained for a variety of lakes large
enough to be effectively imaged at the AVHRR spatial scale and located within the areas of coverage by
CoastWatch imagery. The lakes chosen for this study will include lakes with varying surface area to
depth ratios, salinities, and geographic locations. Examples of target lakes include lakes with large
surface area to depth ratio values (large, shallow lakes) and lakes with small surface area to depth ratio
values (very deep lakes). Large, shallow lakes are typically well mixed and are more likely to be
isothermal over their depth (examples include Oneida Lake New York, Lake Okeechobee Florida, and
the Salton Sea California) than very deep lakes. Very deep lakes are expected to be highly stratified and
heterogeneous in temperature (examples include Pyramid Lake Nevada and Lake Tahoe California). The
comparison of seasonal values of surface temperature derived remotely from AVHRR imagery with
surface and whole water column values will allow for an assessment of the ability of AVHRR imagery to
accurately estimate changes in lake thermal values over time for a variety of lake types.
22
-------
Section 3
Technical Approach
3.1 Work Breakdown Structure
Task A) A time-series of thermal images will be assembled for the 1994 ice-free season for Lake
Superior using NCAAS AVHRR imagery and software and ERDAS IMAGINE 8.3 image
processing software.
Subtask A. 1) Images for the 1994 ice-free season for Lake Superior have been reviewed to
determine images to be included in the image series. Images have been
obtained for all cloud-free and low-cloud image dates. The best images (lowest
cloud cover) were obtained on a minimum of ten-day to two-week intervals to
cover the mid-April to mid-October season.
Subtask A.2) The imagery was decompressed using DECCON software available from
NCAAS), the image files converted from an 11-bit data format to 8-bit data, and
each image registered to the same scale (albedo percentages or Celsius
temperatures). The thermal scale used covers the range of temperatures from
0.1 ° to 25.0 ° C in 0.1 ° increments. This includes the range of surface
temperatures seen in Lake Superior during the season of interest.
The images generated for this project will use the following DECCON
parameters:
Reflective Bands (1 and 2): Conversion to albedo values with a minimum of
0.00 and a maximum of 12.5 units of spectral radiance. These values are the
DECCON default minimum and maximum values for albedo images. Each DN
(Digital Number, the value assigned to a pixel) on the decompressed image
corresponds to a 0.05 unit of spectral radiance. An equal RGB (gray-scale)
image will be created.
Emission bands (3,4,5 MCSST): Conversion to degrees Celsius with a
minimum value of 0.1 and a maximum value of 25.0. DN's of zero will be
reserved for missing data or the masked areas and DN's of 251 - 255 will be
reserved for graphics (overlays can be imbedded in the image if desired). This
will yield a final image where each change of one DN equals a 0.1 degree
Celsius change. The degrees Celsius temperature sensed at a pixel location can
be determined by taking the remotely sensed pixel value (DN) and dividing it
by 10.
23
-------
Subtask A.3) A base mask image was constructed using an *.sc2 (near-IR reflected light)
image acquired in the time series. The mask image was georeferenced to a
polyconic coordinate system. The *.sc2 image used for the masking image for
Lake Superior was the most cloud-free image found that displays the least offset
in the visible bands (the offset is presumed to be atmospheric effect). Pixels
over water were sampled and the DN range 1 7 was determined to be the range
of near-IR values observed for water in that image. Pixels with DN's in the
water range were receded to one and DN's higher than that amount were
assumed land and receded to zero. A NOAA polyconic map (#14960,
1:600,000 scale, 30th edition, 1991), covering all of Lake Superior in a single
map, was used as the base map for georeferencing.
The first reference point selected on the base map was the intersection of the
91.5° W. longitude and 47.0° N. latitude lines (in the west basin near Duluth).
The second reference point was the intersection of the 85.0° W. longitude and
47.0° N. latitude lines (in the east basin near Whitefish Bay). The spheroid
used was WGS 84, the central meridian was set at -84.0 (84° W. longitude), the
axis of origin was 0.0 (the equator), and false easting and false northing were
set to 0.0. The projection accuracy was tested using the digitizing tablet.
Accuracy was within 0.001° longitude/latitude near the reference points and
within 0.003° longitude and 0.002° latitude near Thunder Bay (the farthest
point on the lake from the reference points). Given the large area covered by
the map, this accuracy was accepted and the entry of ground control points
(GCP's) was begun.
Subtask A.4) Eleven ground control points (GCP's) visible at the 1 km2 AVHRR scale,
geographically distributed around the lakeshore, have been selected from the
mask image and entered into an *.gcc file in ERDAS IMAGINE 8.3 (Appendix,
Table Al). Geographically distributed GCP's are needed so that future images
could be registered to the polyconic mask, even if portions of the lake in those
images are obscured by clouds or lost as dropped file sections.
After all the GCP's had been entered, the ERDAS program COORDN was used
to generate the transformation algorithm to be applied to the image. A linear
transformation was used and all 11 GCP's were left in the *.cfn file written.
The RMS error was 2.32 pixels. This is a relatively high degree of error but
was accepted due to the large size of the study area and the desire to retain all
11 GCP's in the *.gcp file. The georectified mask image created in the PC
environment was imported back into the UNIX version of ERDAS (Imagine)
and all 11 GCP's (geographically distributed throughout the lake) were retained.
The georectified mask image will be the base image used for the time series.
All subsequent images will be registered to the same mask by image-to-image
rectification.
Subtask A.5) The thermal MCSST images were imported into Imagine and converted to
*.img images. The thermal images were georegistered to the mask image using
image-to-image registration. A temperature analog color scheme was applied to
the rectified image by using a 256 color, blue-to-red color stretch. A pixel
24
-------
value of one (=0.1 ° C) was assigned a color of pure blue (blue color gun value
of 255, green color gun value of zero, and red color gun value of zero). A pixel
value of 250 (= 25.0 ° C) was assigned a color of pure red (blue color gun value
of zero, green color gun value of zero, and red color gun value of 255). Due to
the color stretch, intermediate pixel values were assigned approximately plus
one additional red color gun value and minus one blue color gun value for each
0.1 ° C increase in temperature. The mask was applied to the *.img image
using the MASK command. In the IMAGINE environment, the latitude and
longitude of each pixel are known and the temperature sensed at any pixel can
be obtained by dividing the pixel DN by 10. In any 8-bit display of the image,
the temperature remotely sensed at any pixel can be determined by sampling the
color values for that pixel. The estimated temperature at that pixel equals [((red
color gun value)/255)*25.0].
Subtask A.6) The ATTRIBUTE EDITOR will be used to assign the color white to pixels in
the masked image that are cloud-covered or that display DN's reduced by thin
cloud cover. A red visible light image (*.scl) and near-IR image (*.sc2)
simultaneously acquired with the MCSST image will be used as a reference
image to show clouds. Clouds will be colored white in the finished base
thermal image to assist in the easy visual identification of cloud-impacted areas.
The use of cloud detection algorithms will be explored to determine if these
algorithms can assist in the determination of cloud-impacted pixels.
Subtask A.7) Using a raster math function created in IMAGINE, cloud-impacted areas of the
base thermal images will be replaced with DN's interpolated from temporally
adjacent image pixels. Reference sets of non-impacted pixels will be used to
scale the interpolation.
Subtask A. 8) Using a raster math function created in IMAGINE, the thermal values
calculated for each water pixel will be summed for a total temperature and
averaged to calculate a mean temperature and variance of the mean. One of the
goals of this research will be to determine if the use of totals or means best
describe the whole-lake changes observed.
Subtask A.9) The whole-lake thermal values calculated for each date will be integrated over
the time series to calculate a single value for the season. A late-fall single data
value will also be calculated. One of the goals of this research will be to
determine if the use of whole season integrated values or just the late-fall heat
content best describe the snowfall amount vs. lake heat content changes
observed.
Task B) The GLSEA dataset for Lake Superior will be assembled for the 1992-1999 period.
Subtask B.I) The yearly thermal values will be compared to buoy data integrated across the
same time and to data from other years. The late fall thermal value will be
compared to buoy data obtained at the same time and to data from other years.
Both values will be compared to snowfall amounts and seasonal temperatures
on time scales appropriate to the data to search for correlation and relationships
for the Lake Superior time series.
25
-------
Task C) Additional lakes and associated data sources to serve as ground truth will be researched.
Surface and vertical temperature data from these lakes will be obtained as well as CoastWatch
imagery for the concurrent periods. The additional lakes will include large inland lakes and the
other Great Lakes covered by CoastWatch imagery. Of particular interest are Lake Michigan,
Lake Ontario, and the lakes studied for ice-cover duration and other long-term research
projects. These lakes may have temperature data available or may provide phenologies that
could be compared to thermal values generated through this research.
3.2 Potential Study Areas
The initial study area for this research project will be Lake Superior. Further study areas will be
researched during the course of this project. If warranted by the results of this research, future research
will extend this technique to additional bodies of water. Additional lakes and associated data sources to
serve as ground truth will be identified for future study. The criteria for inclusion will be based upon the
size of the lake, the availability of thermal imagery over a long time frame, and the existence of other
data sources to act as ground truth.
3.3 Results and Reporting
This study is driven by the hypotheses that:
1) There is a relationship between the mean annual and seasonal surface temperatures that can be
remotely sensed and the annual and seasonal heat content of a large lake; and
2) There is a relationship between the thermal indicator values to be generated and regional
climatology; and
3) There are relationships between the thermal indicator values to be generated and other
phenological measurements such as ice cover duration and lake-effect snowfall.
The criteria for success of this study will be the ability to statistically detect the above relationships.
Results will be reported to the scientific community through a series of internal EPA reports,
symposium proceedings and journal articles in peer-reviewed scientific journals. Results of this research
will be disseminated as appropriate as results warrant.
26
-------
Section 4
The Importance of this Research
to EPA Milestones and Program Area Goals
The EPA, to better account for the success of its actions, has developed a cascading set of goals,
objectives, sub-objectives, milestones, measures, tasks, and products in compliance with the Government
Performance Results Act (GPRA). There are currently ten longer-term goals for the EPA under the
GPRA. Goal 6, "Reduction of Global and Cross-border Environmental Risks" is one of the core areas of
the ORD's Ecological Research Program. The specific objective associated with this research proposal
under the ORD's ecoresearch "Global Risks" goal is 6.2, "Climate Change". The sub-objective under
which this research will be conducted is 6.2.3, "Global Climate Change". Under this sub-objective, "by
2000 and beyond, ORD will provide the capability to assess ecological and associated human health
vulnerability to climate-induced stressors at the regional scale and assess mitigation and adaptation
strategies".
The proposed research advances these goals by studying the potential linkage between climate
change and the impact that lakes have on their surrounding regional environment. More than 40 million
people live around the Great Lakes and the St. Lawrence River valley in areas impacted by lake-effect
snowfall. Increased knowledge about how global climate change may impact lake-effect snowfall will
assist in planning and the mitigation of the economic and health impacts of lake-effect snowfall. The
research techniques proposed in this research project may be applied to a wide set of large lakes in the
Northern Hemisphere. The wide application of this research also may increase our knowledge about the
changes in climate that have occurred over the past 20 years and how these changes relate to other long-
term data sets.
In addition, ORD's "Ecological Research Strategy" identifies major objectives, sub-objectives, and
products associated with its core research program areas of:
1) Ecosystem monitoring research
2) Ecological processes and modeling research
3) Ecological risk assessment research, and
4) Ecosystem risk management restoration research
Shorter-term accounting of success is accomplished by establishing, and monitoring the response to
the annual performance goals (APGs) and measures (APMs) under GPRA and progress toward
completion of any additional critical research products identified in the ORD's "Ecological Research
Strategy" and its subsequent updates. These goals and measures provide the "why" and the "what" of
our research tasks and projects. This document, as a technical research plan addresses not only the
"why" and the "what," but also the "how" »the approach to providing products that satisfy the specific
performance goals associated with this activity.
27
-------
Those specific annual performance goals are:
1) Imagery, Buoy Data, and Meteorological Data Collection
2) Literature Review Report
3) Comparison of values calculated with ground-truth
4) Inter-Agency Progress Reports, Peer-Reviewed Papers and Journal Articles
Those specific annual performance measurements are:
la) All relevant CoastWatch imagery will be obtained and screened for inclusion to the data set by
the end of the first project year. The criterion for inclusion is lake area with less than 40 %
cloud-impacted or missing data values per scene.
Ib) All relevant buoy data will be obtained and screened for inclusion to the data set by the end of
the first project year. The criterion for inclusion is less than 10 % missing data values per
thermal season studied.
Ic) All relevant meteorological data will be obtained and screened for inclusion to the data set by
the end of the first project year. The criterion for inclusion is less than 10 % missing data
values per thermal season studied.
2) A literature review report will be submitted for internal EPA peer review by the end of the first
project year. The criterion for successful completion of this goal is an acceptable internal peer
review (if felt appropriate, this report may be submitted for publication to a peer-reviewed
journal).
3) At the end of the processing of each yearly set of imagery, the temperatures values calculated
via remote sensing will be compared with the ground-truth (buoy) data. The criterion for
successful completion of this goal is a linear regression R2 greater than .9 between the imagery
and buoy values.
4) Inter-agency progress reports, peer-reviewed papers, and journal articles will be submitted for
review and publication on an ongoing basis. The criterion for successful completion of this
goal is acceptance for publication.
28
-------
Section 5
Quality Assurance Statement
This research will be conducted in accordance with the Quality Management Plan for the
Environmental Sciences Division (BSD), National Exposure Research Laboratory, Las Vegas. A Quality
Assurance (QA) Project Plan will be prepared prior to the initiation of field activities and data
processing. The QA Project Plan will document: (1) questions to be answered or decisions to be made
based upon study data; (2) The nature, number and quality of data points needed to achieve a selected
level of confidence in those decisions; (3) the experimental design and methods necessary to meet those
data objectives.
In addition, the Environmental Photographic Interpretation Center (EPIC) has a Master Quality
Assurance Project Plan in place and has developed a full set of 53 Standard Operating Procedures (SOP)
for all aspects of photographic acquisition, scanning, processing, analysis, and graphics. SOP's for the
procedures used in this project will be developed and maintained as a part of the project. Appropriate
SOP's will be referenced in the QA project plan.
29
-------
Section 6
Anticipated Results/Specific Work Products
and Dates for All Sub-tasks
It is anticipated that the result of this research will be an increased knowledge base relating changes
in the water temperature of large lakes and lake-effect snowfall with changes in climate. The specific
work products from this research will be one or more peer-reviewed journal articles. The establishment
of SOP's for the project, the development of the specific QA plan for the project, and the processing of
the initial year of imagery will be completed by 12/00. The acquisition of 1992-1999 years of data and
their processing to yield final products will be accomplished by 12/02.
30
-------
Section 7
Key Milestones
Key Milestones in this research include:
1. Research Plan:
A fully developed and peer-reviewed research plan will be in place by
1/1/00.
2. Literature Review: I will conduct a complete literature review of the remote sensing of thermal
aquatic features, large lake thermal structure, lake effect snowfall, and of
the application of long-term data to climate change study. The initial
literature review will be completed by 1/1/00 and will be ongoing.
3. Field Data Collection: I will collect data over the next two years. Selected meteorological and
limnological data will be collected via remote access to data sets and
external purchase as needed. Data collection will be ongoing.
4. Remote Sensing
Data Collection:
Working with NOAA, GLERL, academic institutions, governmental and
commercial vendors, I will acquire AVHRR thermal imagery over the
longest possible time frame for the study area. Image acquisition will be
ongoing.
5. Research Partnerships: I will be pursue developing cooperative partnerships with Federal Agencies
and academic institutions for the acquisition and analysis of remote sensing
and climate data for environmental issues. I hope that collaborative
research arrangements can be reached with government and non-
government researchers working on similar projects. Partnership
development will be ongoing.
6. Research Results:
Developed on an ongoing basis, I will report research results in the form of
internal reports, symposium papers and peer-reviewed journal articles.
Results will be disseminated as justified and needed.
31
-------
References
Addink, E. A. and A. Stein. 1999. A comparison of conventional and geostatistical methods to replace
clouded pixels in NOAA-AVHRR images. International Journal of Remote Sensing 20(5): 961-977.
Anderson, W. L., D. M. Robertson and J. J. Magnuson. 1996. Evidence of recent wanning and El Nino-
related variations in ice breakup of Wisconsin lakes. Limnology and Oceanography 41(5): 815-821.
Assel, R. A., C. R. Snider and R. Lawrence. 1985. Comparison of 1983 Great Lakes winter weather and
ice conditions with previous years. Monthly Weather Review 113 (3): 291-303.
Assel, R. A. and D. M. Robertson. 1992. Climatic changes near the Great Lakes inferred from 141-year
ice records. Proceedings of the 5th International Meeting on Statistical Climatology, p. 81-85.
Toronto, Canada. June.
Assel, R. A. and D. M. Robertson. 1995. Changes in winter air temperatures near Lake Michigan during
1851-1993 as determined from regional lake-ice records. Limnology and Oceanography 40: 165-176.
Assel, R. A. 1998. The 1997 ENSO event and implication for North American Laurentian Great Lakes
winter severity and ice cover. Geophysical research letters 25 (7): 1031-1033.
Bennett, E. B. 1978. Characteristics of the thermal regime of Lake Superior. Journal of Great Lakes
Research 4: 310-319.
Boden, T. A., P. Kanciruk and M. P. Farrell. 1990. Trends '90: A compendium of data on global change.
ORNL/CDIAC-36. Carbon Dioxide Information Analysis Center.
Bolsenga, S. J. and D. C. Norton. 1993. Great Lakes air temperature trends for land stations, 1901-1987.
Journal of Great Lakes Research 19: 379-388.
Boyce, F. M. 1974. Some aspects of Great Lakes physics of importance to biological and chemical
processes. Journal of the Fisheries Research Board of Canada 31: 689-730.
BrahamR. R., Jr. andM. J. Dungey. 1995. Lake-Effect Snowfall over Lake Michigan. Journal of
Applied Meteorology 34(5): 1009-1019.
Cayula, Jean-F. and P. Cornillon. 1996. Cloud Detection from a Sequence of SST Images. Remote
Sensing of Environment 55(1): 80.
Changnon, S. A. 1997. Climate change and the water levels of the Great Lakes. Shore & Beach 65(2):
4-6.
Chapman, W. L. and J. E. Walsh. 1993. Recent variations of sea ice and air temperatures in high
latitudes. Bulletin of the American Meteorological Society 74: 33-47.
32
-------
Chaston, P. R. 1989. The Magic Chart for forecasting snow amounts. National Weather Digest 14(1)-
20-22.
Croley.T. E. 1990. Laurentian Great Lakes double-CO2 climate change hydrological impacts. Climatic
Change 17: 27-48.
Croley, T. E., F. H. Quinn, K. Kunkel and S. J. Changnon. 1995. Potential Great Lakes hydrology and
lake level impacts resulting from global warming, p. 67-72. Proceedings of the 6th Symposium
Global Change Studies. Dallas TX.
DeStasio, B. T., Jr., D. K. Hill, J. M. Kleinhans, N. P. Nibbelink and J. J. Magnuson. 1996. Potential
effects of global climate change on small north-temperate lakes: Physics, fish, and plankton.
Limnology and Oceanography 41(5): 1136-1149.
Dockus, D. A. 1985. Lake effect snow forecasting in the computer age. National Weather Digest 10(4):
5-19.
Doran, P. T., C. P. McKay, W. P. Adams, M. C. English, R. A. Wharton Jr. and M. A. Meyer. 1996.
Climate forcing and thermal feedback of residual lake-ice covers in the high Arctic. Limnology and
Oceanography 41(5): 839-848.
Eichenlaub, V. L. 1970. Lake effect snowfall to the lee of the Great Lakes: Its role in Michigan.
Bulletin of the American Meteorological Society 51: 403-412.
Eichenlaub, V. L. 1979. Weather and Climate of the Great Lakes Region. The University of Notre
Dame Press. 335 pp.
Eichenlaub, V. L., J. R. Harman, F. V. Nurnberger and H. J. Stolle. 1990. The Climatic Atlas of
Michigan. The University of Notre Dame Press. 165pp.
Ellis, A. W. and D. J. Leathers. 1996. A synoptic climatological approach to the analysis of lake-effect
snowfall: Potential forecasting applications. Weather and Forecasting 11(2): 216-229.
Federal Water Quality Administration (FWQA). 1968. Report on the water movements and temperature
structure in Lake Superior based on 1966-67 field study. U.S. Department of the Interior, FWQA,
Great Lakes Region, Chicago IL. 66 pp.
Fee, E. J., R. E. Hecky, S. E. M. Kasian and D. R. Cruikshank. 1996. Effects of lake size, water clarity,
and climatic variability on mixing depths in Canadian Shield lakes. Limnology and Oceanography
41(5): 912-920.
Goldman, C. R. and A. J. Home. 1983. Limnology. McGraw-Hill, Inc., New York NY.
Gorham, E. and F. M. Boyce. 1989. Influence of lake surface area and depth upon thermal stratification
and the depth of the summer thermocline. Journal of Great Lakes Research 15: 233-245.
Groisman, P. Y., T. R. Karl, and R. W. Knight. 1994. Observed impact of snow cover on the heat
balance and the rise of continental spring temperatures. Science 263: 198-200.
33
-------
Hanson, H. P., C. S. Hanson and B. H. Yoo. 1992. Recent Great Lakes ice trends. Bulletin of the
American Meteorological Society 73: 577-592.
Harrington, J. A., Jr., R. S. Cerveny and K. F. Dewey. 1987. A climatology of mean monthly snowfall
for the conterminous Unites States: Temporal and spatial patterns. Journal of Climate and Applied
Meteorology 26: 897-912
Hobbs, B. F., P. T. Chao, M. Nayal and W. T. Bogart. 1994. Climate warming and Great Lakes
management. Risk Based Decision Making, Water Resources Conference Proceedings, The 6th
Conference on Risk-Based Decision Making in Water Resources; (10/31-11/05/93 : Santa Barbara,
CA, USA) pp. 295-316.
Hoopes, J. A., R. A. Ragotzkie, S. Lien and N. P. Smith. 1973. Circulation patterns in Lake Superior.
University of Wisconsin, Water Resources Center, Technical Report 73-04.
Home, A. J. and C. R. Goldman. 1994. Limnology. McGraw-Hill, Inc., New York NY.
Hutchinson, G. E. 1957. A Treatise on Limnology. I. Geography, Physics, and Chemistry. John Wiley
& Sons Inc., New York. 1015 pp.
Imboden, D. M. and A. Wiiest. 1995. Mixing mechanisms in lakes, p. 83-138. In A. Lerman, D. M.
Imboden, and J. R. Gat [eds.], Physics and chemistry of lakes. 2nd Edition. Springer-Verlag, Berlin.
334 p.
International Joint Commission (JJC). 1977. Physical Limnology (of Lake Superior). The Waters of
Lake Huron and Lake Superior, vol. HI part B: Lake Superior. Upper Lakes Reference Group, pp.
249-275. Windsor, Ontario.
Jassby, A. D., C. R. Goldman, J. E. Reuter and R. C. Richards. 1999. Origins and scale dependence of
temporal variability in the transparency of Lake Tahoe, California-Nevada (USA). Limnology and
Oceanography 44(2): 282-294.
Jiusto, J. E. and M. L. Kaplan. 1972. Snowfall from lake-effect storms. Monthly Weather Review 100:
62-66.
Jiusto, J. E., and H. K. Weickmann, 1973; Types of Snowfall. Bulletin of the American Meteorological
Society 54(11): 1148-1162.
Karl, T. R., R. G. Baldwin and M. G. Burgin. 1989. Time series of regional seasonal averages of
maximum, minimum, and average temperature, and diurnal temperature range across the United
States: 1901-1987. Historical Climatology Series 4-5. National Climatic Data Center.
Keeling, C. D., J. F. S. Chin and T. P. Whorf. 1996. Increased activity of northern vegetation inferred
from atmospheric CO2 measurements. Nature 382: 146-149.
Kevern, N., D. King and R. King. 1996. Lake Classification Systems Pt. 2. The Riparian, Michigan
Lake and Stream Associations, URL: http://www.mlswa.org/lkclassf.htm.
34
-------
Kidwell, K. B., comp. and ed. 1991. NOAA polar orbiter data (TIROS-N, NOAA-6, NOAA-7, NOAA-
8, NOAA-9, NOAA-10, NOAA-11, and NOAA-12) Users Guide. Washington, D.C.,
NOAA/NESDIS.
Kling, G. W. 1988. Comparative transparency, depth of mixing, and stability of stratification in lakes of
Cameroon, West Africa. Limnology and Oceanography 33: 27-40.
Goldman, C. R. and A. J. Home. 1983. Limnology. McGraw-Hill, Inc., New York NY.
Leathers, D. J., T. L. Mote, K. C. Kuivinen, S. McFeeters and D. R. Kluck. 1993. Temporal
characteristics of USA snowfall trends 1945-1946 through to 1984-1985. International Journal of
Climatology 13: 65-76.
Leathers, D. J. and A. W. Ellis. 1996. Synoptic mechanisms associated with snowfall increases to the
lee of Lakes Erie and Ontario. International Journal of Climatology 16(10): 1117-1135.
Lind, O. T. 1985. Handbook of common methods in Limnology, Second Edition. Kendall/Hunt
Publishing Co., Dubuque, IA.
Magnuson, J. J. 1990. Long-term ecological research and the invisible present. BioScience 40(7): 495-
501.
Magnuson, J. J., Wynne, R. H., B. J. Benson and D. M. Robertson. 1999. Lake and river ice as a
powerful indicator of past and present climates. Verhandlungen der Internationalen Vereinigung fur
Theoretische und Angewandte Limnologie. (in press).
Maturi, E. and W. Pichel, 1993. Cloud masking for CoastWatch satellite imagery, oceans 93 Proceedings,
Victoria, British Columbia, Canada, October 18-21, 1993 Volume H, pp. 369-374.
Maturi, E. and W. Pichel, 1994. Cloud Masking for CoastWatch satellite imagery. Proceedings of the 7th
Conference on Satellite Meteorology and Oceanography, June 6-10, 1994, Monterey, CA, American
Meteorological Society, Boston, MA, pp. 569-570.
Mazumder, A., D. J. McQueen and D. R. Lean. 1990. Effects offish and plankton on lake temperature
and mixing depth. Science 247: 312-315.
Mazumder, A., and D. W. Taylor. 1994. Thermal structure of lakes varying in size and water clarity.
Limnology and Oceanography 39: 968-976.
McCormick, M. J. 1990. Potential changes in thermal structure and cycle of Lake Michigan due to
global warming. Transactions of the American Fisheries Society 119: 183-194.
McCormick, M. J. and G. L. Fahnenstiel. 1999. Recent climatic trends in nearshore water temperatures
in the St. Lawrence Great Lakes. Limnology and Oceanography 44(3): 530-540.
Millar, F. G. 1952. Surface temperature of the Great Lakes. Journal of the Fisheries Research Board of
Canada 9: 329-371.
35
-------
Mitchell, J. F. B., S. Manabe, V. Meleshko and T. Tokioka. 1990. Equilibrium climate change and its
implications for the future, p. 131-164. In J. T. Houghton et al. [eds], Climate change: The IPCC
scientific assessment. Cambridge.
Mortimer, C. H. 1993. Long internal waves in lakes: Review of a century of research. University of
Wisconsin - Milwaukee, Center for Great Lakes Studies, Special Report 42.
Mortsch, L. D. and F. H. Quinn. 1996. Climate change scenarios for Great Lakes Basin ecosystem
studies. Limnology and Oceanography 41(5): 903-911.
Mortsch, L. D., S. Quon, L. Craig, B. Mills and B. Wrenn. 1997. Editors. Adapting to Climate Change
and Variability in the Great Lakes-St. Lawrence Basin: Proceedings of a Binational Symposium.
Toronto, Ontario. May 13-15 1997.
Myneni, R. B., C. D. Keeling, C. J. Tucker, G. Asrar and R. R. Nemani. 1997. Increased plant growth in
the northern high latitudes from 1981 to 1991. Nature 386: 698-702.
Namias, J. 1960. Snowfall over the eastern United States: Factors leading to its monthly and seasonal
variations. Weatherwise 13: 238-247.
Niziol, T. A. 1987. Operational forecasting of lake effect snowfall in western and central New York.
Weather and Forecasting 2(6): 310-321.
Niziol, T. A. 1989. Some synoptic and mesoscale interactions in a lake-effect snowstorm. Postscripts,
2nd National Winter Weather Workshop, Raleigh, NC, NOAA, p. 260-269.
Niziol, T. A., W. R. Snyder and J. S. Waldstreicher. 1995. Winter weather forecasting throughout the
Eastern United States. Part IV: Lake effect snow. Weather and Forecasting 10(1): 61-77.
Orlanski, L. 1975. A rational subdivision of scales for atmospheric processes. Bulletin of the American
Meteorological Society 56(5): 527-530.
Phillips, D. W. 1978. Environmental climatology of Lake Superior. Journal of Great Lakes Research 4:
288-309.
Pichel, W., J. Sapper, C. Duda, E. Maturi, K. Jarva and J. Stroup. 1994. CoastWatch Operational
Mapped AVHRR Imagery. International Geoscience and Remote Sensing Symposium, IGARSS'94
August 8-12, 1994. Pasadena, CA, pp. 1219-1221.
Pytlowany, P., comp. and ed. 1994. NOAA Coastal Active Access System (NCAAS) Users Guide.
NOAA Coastal Active Access System (NCAAS) Users Guide. Washington, D.C.,
NOAA/NESDIS/National Oceanographic Data Center, http ://www .node .noaa. gov/NCAAS.
Quinn, F. H. 1998. Potential Effects of Climate Change on the Great Lakes Basin. State of the Great
Lakes, 1997 Annual Report. Michigan Department of Environmental Quality, Lansing MI. pp. 23-
24.
Reycraft, J. and W. Skinner. 1993. Canadian lake ice conditions: An indicator of climate variability.
Climatic Perspectives 15: 9-15.
36
-------
Richards, T. L. and D. G. Massey. 1966. An evaluation of the infra-red thermometer as an airborne
indicator of surface water temperatures. Canada Dept. of Transport, Meteorological Branch,
Technical Circular Series, CIR-4354, TEC-592.
Richards, T. L., J. G. Irbe and D. G. Massey. 1969. Aerial surveys of Great Lakes water temperatures,
April 1966 to March 1968. Climatology Division, Atmospheric Environment Service, Department of
the Environment, Downsview, ON.
Robertson, D. M. 1989. The use of lake water temperature and ice cover as climatic indicators. Ph.D.
thesis, University of Wisconsin-Madison.
Robertson, D. M. and Ragotzkie, R. A. 1990. Changes in the thermal structure of moderate to large
sized lakes in response to changes in air temperature. Aquatic Sciences 52 (4): 360-380.
Robertson, D. M., R. A. Ragotzkie and J. J. Magnuson. 1992. Lake ice records used to detect historical
and future climatic changes. Climatic Change 21: 407-427.
Robinson, D. A. 1991. Merging operational satellite and historical station snow cover data to monitor
climate change. Palaeogeography, Palaeoclimatology, Palaeoecology (Global and Planetary Change
Section) 90: 235-240.
Rothrock, H. J. 1969. An aid in forecasting significant lake effect snows. ESSA Technical
Memorandum WBTM CR-30, National Weather Service Central Region, Kansas City MO. 12 pp.
Schertzer, W. M. and A. M. Sawchuk. 1990. Thermal structure of the lower Great Lakes in a warm
year: Implications for the occurrence of hypolimnion anoxia. Transactions of the American Fisheries
Society 119(2): 195-209.
Schindler, D. W. and multiple others. 1996. The effects of climatic warming on the properties of boreal
lakes and streams at the Experimental Lakes Area, northwestern Ontario. Limnology and
Oceanography 41(5): 1004-1017.
Schneider, K., R. A. Assel and T. E. Croley. 1993. Normal water temperature and ice cover of the
Laurentian Great Lakes: A computer animation, data base and analysis tool. NOAA Technical
Memorandum ERL GLERL-81, NOAA Great Lakes Research Laboratory, Ann Arbor MI.
Schwab, D. J., G. A. Leshkevich and G. C. Muhr. 1992. Satellite measurements of surface water
temperature in the Great Lakes: Great Lakes Coast Watch. Journal of Great Lakes Research 18 (2):
247-258.
Schwab, D. J., G. A. Leshkevich and G. C. Muhr. 1999. Automated mapping of surface water
temperature in the Great Lakes. Journal of Great Lakes Research 25(3): 468-481.
Scott, R. W. and F. A. Huff. 1996. Impacts of the Great Lakes on regional climate conditions. Journal
of Great Lakes Research 22(4): 845-863.
Sloss, P. W. and J. H. Saylor. 1976. Large-scale current measurements in Lake Superior. Technical
Report NOAA ERL 363-GLERL 8,48 pp.
37
-------
Smith, N. and R. A. Ragotzkie. 1970. A comparison of computed and measured currents in Lake
Superior. Proceedings of the 13th Conference on Great Lakes Research 670-675.
Smith N. 1972. Summertime temperatures and circulation patterns in Lake Superior. Ph.D. thesis,
University of Wisconsin-Madison.
Sterner, R. W. 1990. Lake morphometry and light in the surface layer. Canadian Journal of Fisheries
and Aquatic Sciences 47: 687-692.
Vazquez, J., K. Perry and K. Kilpatrick. 1998. NOAA/NASA AVHRR Oceans Pathfinder Sea Surface
Temperature Data Set User's Reference Manual. JPL Publication D-14070, NASA/JPL.
Wetzel, R. G. 1983. Limnology, Second Edition. Saunders College Publishing, Philadelphia, PA.
Wynne, R. H. and T. M. Lillesand. 1993. Satellite observations of lake ice as a climate indicator: Initial
results from statewide monitoring in Wisconsin. Photogrammetric Engineering and Remote Sensing
59(6): 1023-1031.
Wynne, R.H., T.M. Lillesand, F. L. Scarpace and D. C. Rodman. 1995. Metsat-derived lake ice breakup
dates on the Laurentian Shield as a robust climate proxy, p. 123-133. In American Society of
Photogrammetry and Remote Sensing Annual Meeting (61st), Technical Paper V. 2.
Wynne, R. H., J. J. Magnuson, M. K. Clayton, T. M. Lillesand and D. C. Rodman. 1996. Determinants
of temporal coherence in the satellite-derived 1987-1994 ice breakup dates of lakes on the Laurentian
Shield. Limnology and Oceanography 41(5): 832-838.
Wynne, R.H., T.M. Lillesand, M.K. Clayton and J.J. Magnuson, 1998. Satellite monitoring of lake ice
breakup on the Laurentian Shield. (1980-1994) Photogrammetric Engineering & Remote Sensing
64(6): 607-616.
38
-------
Appendix
Lake Superior Thermal Imagery: Aug 16,1994
Polyconic Projection
Pixel Size = 1.1 x 1.1 km
AVHRR MCSST Thermal Imagery
Figure A1. AVHRR thermal image (from CoastWatch MCSST data) of Lake Superior
in mid-August 1994. The surface temperature of the lake exhibits a high
degree of horizontal heterogeneity. Coastal waters in the near-shore regions
of the lake are warmer than water in the central and eastern lake basins.
The Keweenaw Current is apparent along the Wisconsin and Upper
Peninsula shore, and a cold water upwelling is visible along the North Shore
of Minnesota.
39
-------
Table A1. A listing of the 11 GCP's used follows. The longitude/latitude coordinates and the meters
north of the equator and west of 84° W. longitude are listed for each point. These were
determined and generated by the ERDAS Imagine GCP program during a subsequent
georectification process (after the rectified mask image was exported back to the UNIX
system).
GCP #1- St. Louis River mouth
46.75 N lat.; 5209306.25 meters N -92.1 W long.; -616343.125 meters W
GCP #2- Thunder Cape (pt. east of Thunder Bay)
48.2958 N lat.; 5365368.75 m N -88.94166667 W long.; -368155.625 m W
GCP #3- NE tip of Isle Royale (Blake Pt.)
48.190277778 N lat.; 5350243.75 m N -88.42388889 W long.; -331305.625 m W
GCP #4- E tip of Copper Island (far eastern end of Nipigon Bay)
48.76388889 N lat.; 5408406.25 m N -87.35 W long.; -246743.125 m W
GCP #5- NW corner of Montreal Island (in bay along far eastern shore)
47.325 N lat.; 5240243.75 m N -84.7625 W long.; -56718.125 m W
GCP #6- N tip of Caribou Island (east basin)
47.375 N lat.; 5248356.25 m N -85.83333 W long.; -136193.125 m W
GCP #7- Whitefish Point
46.7694444 N lat.; 5180706.25 m N -84.9525 W long.; -75005.625 m W
GCP #8- Manitou Island (off Keweenaw tip)
47.4166667 N lat.; 5262106.25 m N -87.6166667 W long.; -272180.625 m W
GCP #9- Bark Point (SW of Apostle Islands on peninsula)
46.885 N lat.; 5219343.75 m N -91.18694444 W long.; -548143.125 m W
GCP #10- ENE corner of Michipicoten Island
47.74166667 N lat.; 5289743.75 m N -85.594444 W long.; -119005.625 n W
GCP #11- NE corner of Grand Island (Munising, Ml area)
46.55694444 N lat.; 5166681.25 m N -86.65079365 W long.; -203843.125 m W
40
------- |