J. Great Lakes Res. 30(1):184–200
Internat. Assoc. Great Lakes Res., 2004
Temporal Behavior of Levels of the
Great Lakes and Climate Variability
Stanley A. Changnon*
Illinois State Water Survey
2204 Griffith Drive
Champaign, Illinois 61820
ABSTRACT. Levels of Lakes Superior, Michigan-Huron, and Erie for the 1861–2001 period were
assessed to define key temporal fluctuations in their averages and extremes. Behavior of levels of Lakes
Michigan-Huron and Superior has been extremely different since 1861, including vastly different longterm distributions, differences in the amount of variability over time, and differences in when their record
high and low levels occurred. Record high or low 15-year events were present on one or more lakes in 64
years, and record events based on 25-year periods were present in 96 of the 141 years, both representing
the presence of records during much more time than if the record events had occurred simultaneously on
all lakes. These lake level differences reflect significant between-basin differences in climate conditions,
and principally precipitation over time. There were two eras, 1923–1938 and 1973–2001, with exceptional variability and extremes of levels on all lakes. These findings are relevant to planning for future
water level conditions, for understanding recent extremes, and for considering how the sizable spatial
and temporal shifts of the past could relate to future changes in the basin’s climate.
INDEX WORDS:
Lake levels, climate change, climate conditions.
INTRODUCTION
The water levels of the Great Lakes remain one
of the never-ending key issues in the Great Lakes
basin. The levels are constantly changing, and the
fluctuations of the lake levels, including the frequent setting of records for extreme levels and rates
of change, have been a subject of deep concern for
more than 100 years (Changnon 1990). The varied
interests in the lakes over time have become more
sensitive to lake level fluctuations (Horvath et al.
1988). Record high lake levels in the mid-1980s led
the U.S. and Canada to conduct a major water levels reference study, considering possible ways to
ameliorate the impacts of the fluctuations (Water
Science and Technology Board 1989). The extensive damages from the record high levels and ensuing inundation of facilities and the erosion of
shorelines and properties were estimated to range
from $62 million (1992 dollars) to $72 million per
year (Working Committee 2 1993). During
1985–1987 private persons and the governments
spent $102 million on shore protection facilities.
*Corresponding
Low water damages to recreational facilities were
estimated at $1.2 million. Virtually every aspect of
life in the Great Lakes basin is either directly or indirectly affected by fluctuating water levels (Horvath et al. 1988). Thus, those living around the
lakes, and those using the lakes for various purposes including shipping, all have considerable interest, economic and/or environmental, in lake
levels.
Scientific understanding of climatic behavior in
the Great Lakes basin evolved over the past 75
years, revealing that the climate was composed of
short-term (1- to 2-year) fluctuations, mid-term oscillations (2- to 25-year), and longer-term (multidecadal-century scale) trends (Day 1926, Bruce and
Rodgers 1962). In addition to the long-term climate-driven fluctuations, there are three other types
of fluctuations in lake levels. There are short period
fluctuations caused by weather extremes that can
last from a few hours to a few days (e.g., storm
surges and seiches). A second type is the regular
seasonal fluctuation reflecting the annual hydroclimatic cycle from summer high levels to winter
lows. The third type is the fluctuation that results
from the artificial regulation of lake levels by con-
author. E-mail: [email protected]
184
Lake Level Fluctuations and Climate Variability
trol works at the outlets of Lakes Superior and Ontario. The slow crustal movement upwards of the
region also effectively slowly alters lake levels.
The lake levels are a “climate indicator” reflecting both short- and long-term climatic conditions
(Brinkmann 1984). Groundwater inflow to the lakes
reflects historic climatic conditions, whereas
streamflow into the lakes is largely a result of the
climate conditions of the past 3 to 24 months. The
primary climate conditions are the precipitation
over the basin and the evaporation from the lakes
(Brunk 1959). The amount of precipitation on the
lakes and their basins is the single most important
factor, followed by the amount of evaporation,
mainly a wind and temperature-driven variable
(COE 2001). Study of 100-year records of lake levels and precipitation in the Lake Michigan-Huron
basin found a correlation coefficient of +0.74 based
on a 1-year lag. Thus, precipitation explains 55 percent of the variability in the lake levels (Muller et
al. 1965). Evaporation rates reflect temperatures
and airflow over the lakes, and represent 30 percent
of the water lost from Lake Michigan-Huron (COE
1991). Four important physical conditions affecting
levels are not measured and must be estimated, and
these include overlake precipitation, overlake evaporation, groundwater exchange, and overland evapotranspiration (Bruce and Rodgers 1962).
Evaporation, precipitation, and temperature are well
measured over the land area, and runoff is measured
at many but not all streams entering the lakes.
Recent concerns over climate change resulting
from global warming during the 21st Century have
added to the worries over future lake levels
(Changnon and Glantz 1996, Changnon 1997,
Quinn 1999). Will a future change bring lower or
higher levels? Recent shifts in lake levels led to a
major disaster-oriented assessment of the “record”
declines in recent years and attributed these to climate change from global warming (National Geographic 2002). Many scientific studies have
estimated the effects of climate change on future
lake levels (Quinn et al. 1997, Croley et al. 1998).
For these reasons, an assessment of the levels of
the un-regulated middle lakes (Michigan-Huron and
Erie) and their source of inflow (Lake Superior)
was pursued to examine both short-term and midterm historic fluctuations and trends to ascertain
when and where extremes were set.
Lake levels are a result of several physical and
anthropogenic factors with climate the dominant
factor. Humans have made changes in the lakes,
particularly to the channels between them, and
185
these have also affected lake levels. Channels between Lakes Huron and Erie have been deepened
during the 1930s and 1960s, leading to an estimated
lowering of Lake Michigan-Huron by 37 cm over
the past 70 years (Quinn 1999). Such dredging affects the ranks assigned to some of the lake level
extremes. Flow through the channel between Lakes
Superior and Michigan-Huron has been controlled
since 1921 with subsequent changes in the amounts
released in 1946 and 1979. The current outflow regulation, established in the 1977, seeks to bring the
levels of Lakes Michigan-Huron and Superior to
nearly the same relative position within their historic ranges of levels, and specifically to keep Lake
Superior’s levels between 183.4 and 182.3 meters
(COE 1997). However, the capability to regulate the
outflows from Lake Superior does not mean that a
major control of the lake levels is possible. The
major factors (precipitation and evaporation) can
not be controlled. For example, recent lake-level
fluctuations from near record high levels in 1997 to
low levels by 2000 revealed that the Lake Superior
outflow control had only a minimal influence on
levels of Lakes Michigan and Erie (COE 2002).
Other anthropogenic actions have affected levels.
Since the 1870s water has been diverted away from
the Lake Michigan at Chicago with a major increase in the amount diverted in the 1900–1930 period (Changnon and Changnon 1996). This
diversion is now regulated by a U.S. Supreme Court
decree at 90.6 m3/s. Water has also been diverted
into Lake Superior from Hudson Bay basin at Long
Lac (began in 1941) and Ogoki (began in 1943). All
these anthropogenic factors affected lake levels, although the regulation of Lake Superior has produced little bias in the levels of Lake
Michigan-Huron (Quinn and Sellinger 1990). However, the diversion out at Chicago and those in at
Long Lac and Ogoki have only a minor effect on
water levels, being much less than the climate factors affecting lake levels (COE 2000).
Recent decades have experienced record high
levels and sudden severe drops in lake levels, raising two questions: 1) how many years over time actually have one or more lakes had record levels, or
a near record water level, and 2) is the climate
changing? Inspection of the lake levels plotted over
time also suggests there have been distinct periods
with more/less variability in levels, and the variation was assessed over time. A third issue addressed
was the annual rate of change in lake levels, another factor seen as important in understanding lake
levels. How these events reflect the region’s climate
186
Stanley A. Changnon
over time is also addressed. The findings should be
useful to those planning for future lake levels.
DATA AND ANALYSIS
Monthly measures of lake levels for Lakes Superior, Erie, and Michigan-Huron were available from
the U.S. Army Corps of engineers for 1861 to 2001,
a 141-year period. These data were used to calculate 5-, 15-, and 25-year moving averages, and in
turn, to compute their standard deviations. These
three periods were chosen to embrace lengths of
well known short-term fluctuations in continental
climates. The annual mean levels (January-December) were used to compute the rate of change in levels between the average level of a given year and
the average levels of the next 1- and 3-year periods.
These various measures of level changes were computed for the total period of record for each lake.
Climatological data assessed were historical records
collected by Canadian and U.S. weather agencies,
on file at the Midwestern Regional Climate Center,
and the precipitation stations prior to the 1880s are
less dense than those available thereafter.
The selection of times (years) when extremely
high or extremely low levels or variability occurred
on a given lake was based on identifying high periods (or low periods) that were temporally independent without overlapping years. For example, on
Lake Michigan-Huron the maximum 25-year standard deviation value occurred during 1952–1976,
and another high but slightly lesser variable period
occurred during the 1962–1986 period. Since these
periods overlapped, the lesser one (1961–1986) was
not used in the analysis to identify the top three 25year periods of high variability on Lake MichiganHuron during 1861–2001. To save space, mention
of Lake Huron is hereafter deleted since its levels
are the same as those for Lake Michigan.
AVERAGE LAKE LEVELS
The 141-year temporal distributions of the levels
for each lake, based on the 5-, 15-, and 25-year
moving averages, serve as a meaningful base line
for assessing various temporal fluctuations. The 25year averages of Lake Superior levels indicate a
long-term slow rise in lake levels from 1861 until
about 1950 (Fig. 1), with no marked up or down
trend thereafter. Precipitation amounts in the Lake
Superior basin from 1898 to 2001 showed near average values during the 1898–1924 period then
slightly below average values through the 1930s
with above average values from 1940 to 1957
(Changnon 2003). Basin precipitation was again
below average from 1958 to 1972, and this was followed by slightly above average amounts from
1973 to 2001. The fluctuations in levels for the 15year periods (Fig. 1) closely followed these basinwide precipitation fluctuations.
The three sets of average levels for Lake Michigan-Huron (Fig. 2) all display a distinct U-shaped
distribution for the 141-year period. However, the
U-shape is less obvious in the 5-year distribution
than in those for the 15- and 25-year values. Values
for 15- and 25-year periods reached their all-time
lows for periods ending during 1938–1948 period.
Average basin-wide precipitation values for 15-year
periods were near average from 1898 to 1912, but
thereafter were below average until 1956
(Changnon 2003). Amounts during the 1957–1972
period were near average and then became much
above average from 1973 to 2001. This precipitation distribution agrees well with the 15-year lake
level fluctuations (Fig. 2).
The Lake Erie levels had 141-year distributions
(Fig. 3), based on 25-year averages, similar to that
for Lake Michigan-Huron. The lowest values occurred during the 1931–1945 period, in close agreement with the Lake Michigan distribution.
However, the early period high values on Lake Erie
were less than those in recent years, a notable difference from the magnitude of the early and late
levels of Lake Michigan.
The temporal distribution of the Lake Superior
levels for 15-year periods (Fig. 4) is markedly different from those for the other two lakes. It peaks in
1907 and 1940 when the other lakes were low, and
the 141-year trend is very different. These differences in the average distribution for the three lakes
reveal that the climatic conditions affecting Superior ’s basin from 1861 until about 1950 were
markedly different from those affecting MichiganHuron and Erie. A study of precipitation over each
lake’s basin for the 1870–1979 period found that
the major decrease in levels of Lake MichiganHuron during the 1870–1940 period either was not
present or was very minor in the other lake basins
(Quinn and Croley 1981).
A climatological study of the spatial frequency of
cyclones, which produce much of the precipitation
in the Great Lakes basin, showed a sharp northsouth gradient in frequencies across the basin
(Angel 1996). The average annual number of cyclones over the Lake Superior basin ranges from 31
to 35, whereas the averages across southern Lake
Lake Level Fluctuations and Climate Variability
187
FIG. 1. Average lake level curves based on the 5-, 15-, and 25-year moving averages of the levels of Lake
Superior for 1861–2001.
Michigan and Lake Erie range from 25 to 29 cyclones per year (Changnon et al. 1995).
The upward trends of the levels of Lakes Michigan and Erie since about 1945 (Fig. 4) reflect an upward trend in annual precipitation across their
basins and most of the nation since 1935–1940
(Karl and Knight 1998). The national pattern of precipitation trends showed increases in the basins of
Lakes Michigan and Erie, but no upward trend in
the Lake Superior basin. This helps explain the lack
of a comparable increase in levels of Lake Superior.
The regulation of Lake Superior levels also tended
to maintain levels in the upper portion of the distribution (Quinn 1978).These precipitation increases
have led to marked increases from 1948 to 1998 in
annual streamflows in Wisconsin, lower Michigan,
and Ohio where rivers and streams feed Lakes
Michigan and Erie (Lettenmaier et al. 1994). Their
analysis of national temperature trends showed significant increases since 1948 in northern Wisconsin
and Minnesota, including the basin of Lake Superior, and this may relate to an increase in lake evaporation, another factor limiting any increase in the
lake’s levels. The lower Michigan and Ohio area
had significant temperature decreases for
1948–1998, a condition which would likely reduce
evaporation over time and also lead to increased
lake levels. One analysis of the fluctuations in levels of the lakes claimed a distinct basin-wide
change in climate occurred around 1970 (Hartmann
188
Stanley A. Changnon
FIG. 2. Average lake level curves based on the 5-, 15-, and 25-year moving averages of the levels of
Lakes Michigan-Huron for 1861–2001.
1988), whereas another assessment claimed the
1900–1986 period had two distinct precipitation
patterns: one classed as below average (basin mean
of 78.2 cm) for 1900–1940, and one with above average amounts (mean of 83.2 cm) for the
1941–1986 period (Great Lakes Commission 1986).
RECORD HIGH AND LOW LEVELS
The ending years of the three highest and three
lowest levels for 5-year periods attained on each
lake during 1861–2001 appear on Table 1. The values show considerable temporal grouping of the
high and low periods. For example, two lakes had
high levels in the early 1970s, ranked as the second
highest of the 141-year period. All three lakes had a
peak again during the early 1980s. However, four of
the nine highest values were isolated in time and on
a single lake. For example, only Lake Michigan had
a high during 1883–1887, and only Lake Superior
had a high during 1950–1954.
Timing of the low periods on Lakes Michigan
and Erie were in close agreement, all ending in
1927, 1936–1937, and 1966. This reflects the strong
impact of the outflow from Michigan to Erie, as
well as similarity in their climate conditions
(Brinkmann 1983a). Two of the low-level periods
Lake Level Fluctuations and Climate Variability
189
FIG. 3. Average lake level curves based on the 5-, 15-, and 25-year moving averages of the levels of Lake
Erie for 1861–2001.
on Lake Superior occurred during the 1864-1893
period, much earlier than those on the other two
lakes. This outcome reveals that weather conditions
capable of producing extremely dry conditions over
relatively short periods, such as five years or less,
often exhibit considerable spatial continuity across
the basins of Lakes Erie and Michigan. But, these
anomalous conditions seldom extend over the Lake
Superior basin, and those over the Lake Superior
basin often do not extend over Lake Michigan
(Brinkmann 1983b).
The years of occurrence of the first and second
ranked high 15-year levels, and the first and second
ranked low levels for 15-year periods for the three
lakes are shown in Table 2. The impact of the high
and low 5-year levels (Table 1) is evident in the 15year outcomes. Lakes Michigan and Erie both had
low 15-year periods ending in 1937–1938 and in
1969, and both periods embraced the low 5-year periods ending in 1936 and 1966 (Table 1).
Analysis based solely on the record highest and
lowest 15-year periods reveals that 77 years during
1861–2001 were without a record on one of the
three lakes, whereas 64 years (45 percent of the
total) experienced a record event on at least one of
the three lakes. If the three record highs and lows
had all occurred at the same time, as occurred in
1955–1969 for Lakes Michigan and Erie (Table 2),
190
FIG. 4.
Stanley A. Changnon
Moving average lake levels based on 15-year periods.
records would have occurred in only 30 of the 141
years, not in the 64 years with records. In essence,
several extreme lake levels were occurring quite independently such as a record low on one lake and
none on the other two, leading to many more years
than expected with record levels.
An analysis of the extreme lake levels based on
values for the 25-year periods resulted in the periods shown in Table 3. During the 1861–2001 period, there were 45 years when no record level was
occurring, but a record high and/or low occurred in
96 of the 141 years, or 69% of the total time. If
there had been timing agreement of the three highest and three lowest periods of all three lakes, only
50 years would have experienced a record. The
main temporal overlap in record periods was in the
low periods on Lakes Michigan and Erie (both had
lows during 1921–1942), and the high on Lake Superior overlapped them from 1928 to 1945.
The findings showing the periods when record, or
near record, high and low lake levels occurred,
based on 15- and 25-year periods, reveal that considerable spatial difference frequently existed between the lakes with record highs on one lake when
record lows occurred on another lake. This reveals
that major spatial variations in both precipitation
and evaporation exist across the three basins for
such multi-year periods. The results also reveal that
the influence of these interconnected lakes on their
levels is not always large. Record levels (high or
Lake Level Fluctuations and Climate Variability
191
TABLE 1. Five-year periods when extremely high
and extremely low lake levels occurred on Lakes
Michigan, Erie, and Superior. These include the
three highest and three lowest ranked values during
1861–2001. Shown for each period is the lake (M =
Michigan, S = Superior, E = Erie), and the rank of
three periods with 1 = the most extreme and 3 = the
third ranked extreme value.
TABLE 3. The 25-year periods with the record
highest and record lowest lake levels.
Ending year of highest
three 5-yr value
1878 (M-2)
1887 (M-1)
1954 (S-3)
1975 (S-2)
1976 (E-2)
1987 (S-1, M-3, E-1)
1998 (E-3)
The general disagreement in the timing of the
record high and low levels on Lakes Michigan and
Superior is not surprising. The correlation coefficient between their annual basin precipitation
amounts for 1860–1989 period was only +0.62, indicating that annual precipitation on Lake Superior
explained only 38% of the variations in annual precipitation over Lake Michigan basin.
Assessment of 12-year wet and dry periods on
the two basins, defined as periods having at least 8
or more years with above/below average values,
was used to define the occurrence of a given dry or
wet condition on Lake Superior when one existed
on Lake Michigan. As shown in Table 4. when
Lake Michigan experienced a dry period during the
1854–1927 era, one also existed on Lake Superior
in 44 percent of the years, but in 13 percent of the
dry years on Lake Michigan, there was a wet period
on Lake Superior, and in 43 percent of the years
neither wet or dry conditions existed in the Lake
Superior basin. A stronger relationship of dry conditions existed in the later (1928–2000) period.
When wet periods existed on Lake Michigan, wet
periods on Lake Superior were fewer than found for
joint dry periods. For example, when wet conditions occurred on Lake Michigan, only 34 percent
(1854–1927) were wet on Lake Superior, and in the
1928–2000 period this relationship was 39 percent.
Both sets of relationships were slightly stronger in
the second era, reflecting a temporal shift in regional climate conditions.
Ending year of lowest
three 5-yr values
1869 (S-1)
1893 (S-3)
1926 (S-2)
1927 (M-3, E-2)
1936 (E-1)
1937 (M-1)
1966 (M-2, E-3)
TABLE 2. The years when the two highest
ranked lake levels and two lowest ranked lake levels occurred, based on 15-year periods.
Highest two levels1
1874–1888 M-1
1876–1890 E-2
1938–1952 S-1
1972–1986 S-2
1973–1987 E-1, M–2
1 M = Michigan, S = Superior,
= rank two highest level.
Lowest two levels
1863–1877 S-2
1878–1892 S-1
1923–1937 E-1
1924–1938 M-1
1955-1969 M-2, E-2
E = Erie; 1 = rank one, 2
low) for 15-year periods were present in 45 percent
of all years sampled and in 69 percent of the years
with 25-year extremes. If the analysis of 15-year
periods includes the two highest and two lowest independent periods on each lake during 1861-2001,
these accounted for 66 percent of all years from
1861 to 2001.
Years of highest value
1865–1889 Michigan
1928–1953 Superior
1963–1997 Erie
Years of lowest value
1865–1889 Superior
1918–1942 Erie
1921–1945 Michigan
TABLE 4. The frequency of 12-year wet and dry periods on Lake Superior
when either condition existed on Lake Michigan for early and late eras,
expressed as a percentage of the period.
Dry periods on
Lake Michigan
1854–1927 era
1928–2000 era
Lake Superior
had a Dry period
44%
64%
Lake Superior
had a Wet period
13%
6%
Lake Superior had no
Wet or Dry period
43%
30%
Wet periods on
Lake Michigan
1854–1927 era
1928–2000 era
Lake Superior
had a Dry period
25%
30%
Lake Superior
had a Wet period
34%
39%
Lake Superior had no
Wet or Dry period
41%
31%
192
Stanley A. Changnon
VARIABILITY OF
LAKE LEVELS OVER TIME
The standard deviations (SD) of the lake levels
were calculated for the moving averages of each of
the three periods (5, 15, and 25 years). These SD
values were used as a measure of the temporal variability in lake levels between 1861 and 2001.
The standard deviation values on Lake Michigan
for 5-year periods were the highest, and those of
Lake Superior were the lowest throughout the 141year period. The magnitude of the fluctuations in
the SD values of Lake Superior did not shift during
the 141-year period of study. The lack of an increase since the 1920s, as found on the other lakes,
may reflect the regulation of Superior’s outflows
that began in 1921. Lake Erie’s SD fluctuations
from 1861 until the 1920s were different from those
of Lake Michigan. Thereafter, the timing of the
highs and lows in SD values on Lake Erie were
similar to those on Lake Michigan, and levels of
both lakes had greater variability after 1920.
All three lakes experienced four major highs in
their 5-year SD values. The one during 1925–1933
exhibited the greatest variability on Superior but the
variability peaked during 1986–2001 period on the
other two lakes. The periods with extreme variability in lake levels for 5-year periods are listed in
Table 5. Levels on all three lakes exhibited extreme
variability during the 1925–1934 era, a time when
the nation’s major droughts of the 1930s began. All
three lakes also had high variability values for 5year periods during the 1996–2001 era. The other
three periods of high variability occurred as timeisolated events on only one lake.
The SD values calculated for the 15-year moving
averages of the three lakes from 1861 to 2001 (Fig.
5) reveal several between-lake differences. First,
the 141-year variability of levels of Lake Superior
with SD’s of 0.45 to 0.7 was notably less than those
for Lakes Michigan-Huron and Erie with SD’s of
0.6 to 1.5. Regulation of Lake Superior’s outflow to
Lake Michigan began at minor level in 1902, then
was altered in 1921 to a control plan established in
1916, an endeavor not closely followed until 1941
when a new control plan was issued (COE 1993).
From 1941 to present, different regulatory rules
have been in place and closely followed, and a regulation plan of 1977 sought to deal specifically with
high lake levels and to balance levels between Lake
Michigan-Huron and Lake Superior. Comparison of
the SD values for Lake Superior before 1921 and
those after 1941 does not suggest changes in the de-
TABLE 5. The three 5-year periods on each lake
with the greatest variability in the lake levels.
Rank
1
2
3
Lake Superior
1925–1929
1996–2000
1869–1873
Lake Michigan
1997–2001
1929–1933
1949–1953
Lake Erie
1997–2001
1930–1934
1986–1990
gree of variability with time, either before or after
the different regulations.
A second finding derived from Figure 5 is that
the variability on Lake Erie had a similar time distribution as did the values for Lake Michigan, but
the Lake Erie SD values were mostly lower. Variability of levels of both lakes peaked in the
1960–1975 era, whereas the peak SD value on Lake
Superior was during 1925–1939.
The 15-year SD values of Lakes Michigan and
Erie generally increased over time from 1861 to
1975, and thereafter became lower. The SD values
for Lake Superior showed no up trend and were essentially flat for 141 years. Variability values of
Lakes Superior and Michigan were similar during
1900–1920 when both basins had near average precipitation (Changnon 2003), but the magnitude and
type of fluctuations differed in all other years.
Figure 6 presents the ranges in precipitation values, expressed as percent of average, for seven discrete 15-year periods on the basins of Lakes
Superior and Michigan for the 1898–2002 period
(Changnon 2003). The ranges shown are based on
the highest and lowest values from the numerous
climate districts in each basin. Comparison of the
curves of the two lakes reveals two important findings relevant to the SD values of the lake levels.
First, comparison of the two lakes’ sets of curves
clearly illustrates that Lake Superior experienced
much less variability of precipitation in all but one
of seven 15-year periods, thus agreeing with findings exhibited on Figure 5. Second, the variability
of precipitation amounts on the Lake Michigan
basin began a major increase after 1942 that easily
outmatched that on Lake Superior. Variability of
precipitation on the Lake Superior basin raised
from that during the droughts of the 1928–1942 period to 1943–1957, but the variability was small and
as shown on Figure 6, did not change much in the
following three 15-year periods. This could reflect
regulation endeavors.
Comparison of the SD values based on the 25year averages showed Lake Michigan’s values were
highest when Superior’s values were lowest. Fur-
Lake Level Fluctuations and Climate Variability
FIG. 5.
193
Curves of the variability in lake levels for 15-year periods on all lakes.
thermore, the general long-term trends of the three
lakes differed. SD values for Lakes Michigan and
Erie exhibited an upward trend from 1861 to 1988,
whereas the trend of the 141-year SD values on
Lake Superior was essentially unchanging with
time. The distributions of 25-year variability values
on Lakes Michigan and Superior were unlike for
the 1861–2001 period. For example, Lake Superior
had a peak in SD values during the 1921–1945 period and this was a low variability period on Lake
Michigan. Lake Michigan’s all-time peak occurred
during 1950–1974, whereas that on Lake Erie oc-
curred during 1963–1987. The SD values for Lakes
Michigan and Erie dropped rapidly after 1985.
The times when periods of high variability of
lake levels occurred were assessed for each lake
and findings intercompared for the three durations
(5, 15, and 25 years). The three periods on each
lake with the highest SD values were identified and
those selected had to be independent in time with
no overlap in years.
The three periods with the highest variability in
lake levels for the 15-year and 25-year periods appear on Figure 7. Four periods are shown for Lake
Erie since the third and fourth ranked periods had
194
Stanley A. Changnon
FIG. 6. Range of precipitation values, expressed as a percent of average, on the basins of Lakes
Michigan and Superior for seven 15-year periods during 1898–2002.
essentially equal values. The top variability for 25year periods shows that all three lakes were experiencing high variability during three time periods:
1884–1900, 1928–1937, and 1972–1987. The ranks
of the 15-year period values displayed on Figure 7
show little temporal agreement. For example, each
of the three lakes had highly variable 15-year periods ending during the 1931–1936 era, but their
ranks were #1 on Superior, #2 on Erie, and #3 on
Michigan. Inspection of the 25-year variability periods reveals Lakes Michigan and Erie experienced
their greatest variability during the late 1960s to
mid-1980s, but Lake Superior’s variability in this
same period was only ranked third highest during
the 141-year period.
MAJOR LEVEL CHANGES DURING
1- AND 3-YEAR PERIODS
The average annual levels on each lake were
compared with the averages occurring during the
next year and ensuing 3 years. The aim was to assess the major short-term shifts, up and down, in
lake levels. Relatively rapid major changes in levels
often create major adjustment problems for many
lake activities (Horvath et al. 1988).
The shifts in levels between individual years on
Lake Michigan are shown on Figure 8. The shift
from 1930 to 1931 was a decrease to 53.3 cm below
average, the greatest 1-year decrease during the
1861–2001 period. Inspection of Figure 8 also
shows other large singular 1-year increases and de-
Lake Level Fluctuations and Climate Variability
195
FIG. 7. The three periods with the greatest variability in lake levels during 1861–2001 for three lakes
and for 15- and 25-year periods.
creases, and in general, the changes since 1920
have been greater than those in the prior 60 years.
The top four rated single year increases and decreases during 1861–2001 for the three lakes are
listed in Table 6. Assessment of the years of occurrence shows that several periods of decreases extended across all three lakes including 1987–1988
and 1998–1999, with Lakes Michigan and Erie experiencing a major drop (below average) in 1931.
None of the years of major 1-year increases were
found on all three lakes, although an increase in
1929 occurred on two lakes and one in 1996–1997
on two lakes. These results reveal that when extremely dry conditions of 1- to 2-year durations occurred, they extended over more of the basin than
did extremely wet conditions. Eight years with
major increases were events found on only one
lake. The years of the decreases show very few occurred before 1930 (only in 1917 on Lake Superior), and 6 of the 12 major decreases occurred
during the 1987–1999 era. The years when increases
occurred showed a greater temporal distribution,
occurring in the period from 1876 to 1997. The decreases for a given rank on Lake Superior were ap-
proximately 35 to 50 percent less than those on
Lake Michigan.
Changes in lake levels from the annual average
in a given year and the average level of the ensuing
3 years also were assessed. The annual changes on
Lake Michigan for 1861 to 2001 revealed an overall
increase in the magnitude and frequency of extreme
changes after the mid-1920s. The peak 3-year increase was 33.8 cm above the average level from
1926 to 1927–1929, and the greatest decrease occurred from 1929 to 1930–1932 (Table 7).
Comparison of Lake Superior’s temporal distribution with that for Lake Michigan revealed major
differences. The pre-1930 years on Superior had
changes comparable in magnitude and frequency to
those in the post-1930 period, whereas those on
Lake Michigan increased after 1930. The results do
not suggest that regulated outflows from Lake Superior had any influence over time on the magnitude of the major 3-year shifts in lake levels. The
magnitudes of the two lakes’ values for 1997 to the
1998–2000 period are not the same, with Lake
Michigan having a decrease from 1997 to
196
Stanley A. Changnon
FIG. 8. The magnitude of changes in lake levels between consecutive years for 1865–2001 on Lake
Michigan-Huron. Values are meters above or below the long-term average.
1998–2000 of 29.9 cm, double the 14.0 cm drop on
Lake Superior (Table 7).
Inspection of the major 3-year shifts on all three
lakes (Table 7) reveals that three time periods had
major shifts on all lakes. Major 3-lake decreases occurred during 1930 –1933, 1986–1989, and
1998–2001. Major 3-lake increases occurred during
1926–1929, 1949–1953, and 1965–1969. The major
feature on all three lakes was the big swing from
top ranked increases in the late 1920s to large decreases in the early 1930s.
SUMMARY AND CONCLUSIONS
Water levels of the Great Lakes remain a major
issue. The fluctuations of the lake levels, and the
frequent occurrence of extreme levels and rates of
change, have long been a subject of deep concern.
These fluctuations largely reflect shifting climate
conditions over time since climate is the major factor controlling lake levels. This assessment of the
levels of four upper lakes (Superior, MichiganHuron, and Erie) included examining short-term
TABLE 6. The four greatest increases and decreases in lake levels between single years, measured in
centimeters, from average lake levels, and year of occurrence, and during the 1861–2001 period.
Lake Michigan
Decreases
53.3–1931
47.9–1999
40.5–1988
39.3–1977
Increases
47.8–1951
47.2–1960
45.1–1929
35.1–1952
Lake Superior
Decreases
27.4–1998
26.5–1987
18.6–1917
17.4–1940
Increases
38.1–1927
30.5–1916
20.7–1894
20.6–1996
Lake Erie
Decreases
56.4–1931
46.0–1999
40.5–1988
34.4–1934
Increases
42.4–1876
36.6–1943
35.4–1929
35.1–1997
Lake Level Fluctuations and Climate Variability
197
TABLE 7. The five greatest increases and decreases in lake levels between single years and the average
level of the ensuing three years, measured in centimeters, and years of occurrence, and for the 1861–2001
period.
Lake Michigan
Decreases
Increases
31.7:1930–32
33.8:1927–29
29.9:1987–89
29.3:1950–52
28.0:1998–00
22.6:1951–53
26.2:1931–33
20.7:1965–67
25.6:1999–01
18.0:1967–69
Lake Superior
Decreases
Increases
14.0:1998–00
18.6:1927–29
11.6:1986–88
14.6:1926–28
11.0:1987–89
11.0:1892–94
10.4:1877–79
10.7:1893–95
9.8:1952–54
10.4:1949–51
and long-term fluctuations and trends since 1860 to
ascertain when and where extremes occurred and
how the variability of levels has shifted over time.
The 141-year average lake level distributions for
5-year and longer periods show a U-shaped temporal distribution for Lake Michigan, being highest
early (19th Century). The distribution for Lake Erie
was also U-shaped but the highest values came in
recent years (1970–2001), and that for Lake Superior shows a very different distribution. It exhibited
a gradual increase over time from 1861 until about
1950 and a flat trend thereafter.
A review of the fluctuations in the levels of these
upper lake systems offered an interesting perspective on the occurrence of record high or low lake
levels. Depending on the level criteria being examined, record highs or lows were being set on one or
more lakes during nearly half the time over the 141year period. “Records are set to be broken” is an
old adage that appears appropriate for the levels of
the Great Lakes. Assessment of extremes in lake
levels, high or low, shows that record extremes for
15-year periods were occurring on one or more
lakes during 45 percent of the years since 1860, and
25-year records (high and/or low) prevailed on 69
percent of all 141 years. Certain periods including
1884–1900, 1928–1937, and 1972–1987 had considerably more year-to-year variability in lake levels
than did other years, a condition reflecting times of
greater climate instabilities in the Great Lakes
basin. Recent (1972–1987) high levels were records
for Lakes Michigan and Erie, but not for Lake Superior. Major short-term, 1- and 3-year, changes in
lake levels revealed sizable differences between occurrences on the three lakes. Major differences exist
in the timing of record shifts between the 1-year
and 3-year changes. The considerable differences
found in the behavior of lake levels during the
1861–2001 period explain the frequent setting of
record (or near record) high and low levels. This
Lake Erie
Decreases
Increases
23.8:1998–00
23.5:1927–29
21.9:1987–89
17.4:1950–52
21.3:1999–01
17.1:1966–68
20.1:1931–33
16.8:1971–73
19.2:1930–32
16.5:1967–69
type of behavior was driven by considerable temporal and spatial variability in the regional climate
conditions that control lake levels (precipitation and
evaporation).
The variability in the levels of Lakes Michigan
and Erie was less during 1861–1930 than during the
1931–2001 period when it gradually increased over
time. However, the magnitude of the variability in
the level of Lake Superior did not change during
the 141-year period. There were two periods when
the levels of all three lakes exhibited great variability: 1925–1936 and 1972–1989, illustrating the existence of highly variable, less stable atmospheric
conditions. Both periods had a mix of exceptionally
wet and dry years.
A climatological interpretation of the historical
fluctuations in lake levels is potentially limited by
anthropogenic changes in the channels between the
lakes. Alterations in the control of outflows from
Lake Superior to Lake Michigan could conceivably
alter record levels, the degree of variability, and
rates of level changes from one year to another.
However, inspection of the long-term behavior of
the levels of Lake Superior does not suggest a detectable signal relating to the outflow changes. The
temporal distributions of extremely high and low
levels, and the variability of levels of Lake Superior
do not appear to reflect changes in the controlled
outflow since 1921. Effects of the controlled outflows on Lake Michigan levels are not evident either. For example, after the shift to strict controls of
the outflows in 1941, Lake Michigan levels, as
measured for 5-, 15-, and 25-year periods, did not
show post-1941 perturbations in levels. Furthermore, 1941 was followed by ever increasing levels
(but not reaching the high levels of the 19 th Century).
What do these comparisons of averages, extremes, and variability of levels of the four lakes reveal about the climate conditions of their basins
198
Stanley A. Changnon
over the past 141 years? Recall, of course, that two
of the critically important conditions affecting lake
levels go unmeasured including overlake precipitation and evaporation, and their impact on levels is
best revealed in how the levels behave. The comparisons of the levels of the lakes revealed two important findings about the past climate conditions
that control lake levels.
First is the strong evidence of two major periods
of change in the climate that significantly affected
levels of all four lakes. One period occurred during
the 1923–1938 era:
• The variability of the levels of all lakes was exceptionally high during these 16 years.
• The levels reached lows in 1923–1926, rose
rapidly (1927–1929), fell rapidly (1930–1932),
and reached record lowest levels in the 1930s.
• After this period the levels of three lakes increased (but not Superior).
• After this period the variability of three lakes
was much greater than in prior years (but not
Superior).
The second period of instability occurred during
the 1973–2001 period:
• The variability of levels on all four lakes was
exceptionally large (in 1973–1987 and
1996–2001), but did not match values during
the 1923–1938 era.
• All four lakes had record high, or near record
highs twice during 1973–1987 with 1982–1987
the highest period on all lakes.
• Lake levels oscillated, descending rapidly
twice: during 1987–1989 and 1998–2001
(greatest fall during the 141-year period).
Other studies have noted there had been some
form of climate change on the basin during the 141year period. One was reported to have occurred
around 1970 (Hartmann 1988), and the Great Lakes
Commission (1986) noted a change in annual precipitation around 1940. Another study that compared precipitation on the basins of Lakes Superior
and Michigan-Huron concluded that a major change
had occurred during the late 1920s (Changnon et al.
1990). These various observations had identified
one of the two periods of change noted herein.
The second important climate-related finding is
the clear evidence of major spatial differences in
the climates of the basins of the lakes. Climate conditions controlling lake levels on Lake Michigan-
Huron are quite different than those controlling levels of Lake Superior:
• Their average height distributions for
1861–2001 are totally different.
• Times of most of their near record highs and
lows differ, particularly for the high levels, although agreement was reached in the two exceptional climate periods (1923–1938, and
1973–2001).
• The variability of levels on Lake MichiganHuron (for all durations) is much higher than
that for Lake Superior levels, with the variability of Michigan levels increasing steadily after
1936 but not the Superior levels.
These differences in the behavior of the lake levels and hence climate conditions, have been noted
by others. Brinkmann (1983b) found a distinct climate difference between the two basins, and another study found major temporal disagreements in
the incidence of 12-year wet and dry periods
(Changnon et al. 1990). The correlation coefficient
between the precipitation of the two lakes is only
+0.62. Angel (1996) found differences in the temporal distribution of precipitation-producing cyclones with trends upward over time on
Michigan-Huron and Erie basins, but no up or
down trends in cyclone frequency over the Lake
Superior basin.
Due to the large influence of the outflow from
Lakes Michigan-Huron on the levels of Lake Erie, a
good interaction is expected and many aspects of
the levels of Lakes Michigan-Huron and Erie were
alike. This included their average long-term distributions (flat then increasing after 1935), their joint
increases in variability of levels after 1935, and in
the times of occurrence of their record high and low
levels. However, the levels of the two lakes disagreed in three ways. The highest levels on Michigan-Huron were in the 1860–1890 period but the
highest on Erie came during 1970–2000. This is
partly a result of the dredging of the channels between the two lakes (Quinn 1999). Furthermore, the
high variability of levels and the highest levels attained for 25-year moving average periods were not
in agreement.
Examination of the extremes and fluctuations in
the lake levels for 5-year periods showed more
agreement amongst the lakes than found in their
levels for longer, 15- and 25-year periods. This reveals that climate conditions affecting the lake levels, precipitation and evaporation, are more
Lake Level Fluctuations and Climate Variability
frequently similar across all four basins for shortterm periods, such as 5 years, but the conditions
differ much more when sampled over 15-year or
longer periods.
The 1925–1936 period of great variability and
record low lake levels was followed by a shift in
the climate on three lakes (Michigan-Huron and
Erie). This has been noted to have been an ever
wetter period (Karl and Knight 1998) and a cooler
period (Lettenmaier et al. 1994) over their basins,
conditions conducive to increased lake levels. On
Lake Superior’s basin the temporal precipitation increase did not occur and temperatures rose over the
last four decades, conditions that would not result
in lake level increases as occurred on the other
lakes. The key question relates to what type of climate will follow the unstable conditions of
1973–2001?
The results should have considerable value for
those planning for future water level episodes.
Those attempting to understand recent major shifts
in lake levels can view these conditions from a long
historical perspective based on climate aberrations
sampled over 140 years. The findings also reveal
major climate fluctuations, both in space across the
basin and over time, factors very relevant to considering the dimensions of potential future climate
changes and their effects on lake levels.
ACKNOWLEDGMENTS
The lake level data were provided by the U.S.
Army Corps of Engineers Office in Detroit, and I
appreciate their assistance. Jon Burroughs prepared
the graphics.
REFERENCES
Angel, J.R. 1996. Cyclone Climatology of the Great
Lakes. Illinois State Water Survey, Champaign, IL,
Miscellaneous Pub. 172.
Brinkmann, W.A.R. 1983a. Secular variations of surface
temperature and precipitation patterns over the Great
Lakes region. J. Climatol. 3:167–178.
———. 1983b. Association between net basin supplies to
lake Superior and supplies to the lower Great Lakes.
J. Great Lakes Res. 9:32–45.
———. 1984. Association of summer temperature and
precipitation patterns over the Great Lakes region and
water supplies in the Great Lakes. J. Climat.
4:182–188.
Bruce, J.P., and Rodgers, G.K. 1962. Water Balance of
the Great Lakes. GLI7, Great lakes Institute, University of Toronto, Toronto.
199
Brunk, I. 1959. Precipitation and the levels of Lakes
Michigan and Huron. J. Geophys. Res. 64:1591–1595.
Changnon, D., Noel, J., and Maze, L. 1995. Determining
the average cyclone frequencies using equal-area circles. Monthly Weather Review 123:2285–2294.
Changnon, S.A. 1990. Great Lakes waters: Too little or
too much? In Proc.Supplying Water and Saving the
Environment for Six Billion People, pp. 41–49, Amer.
Soc. Civil Eng., Washington, D.C.
———. 1997. Climate change and the water levels of the
Great Lakes, Shore and Beach 65:4–6.
———. 2003. Temporal Distribution of Midwestern Precipitation During the 20 th Century. Illinois State
Water Survey, Champaign, IL.
———, and Changnon, J.M.1996. History of the Chicago
diversion and future implications. J. Great Lakes Res.
22:100–118.
———, and Glantz, M.L. 1996. The Great Lakes diversion at Chicago and its implications for climate
change. Climatic Change 32:199–214.
———, Leffler, S., and Shealy, R. 1990. Effects of Past
Low Lake Levels and Future Climate-related Low
Lake Levels on Lake Michigan, Chicago, and the Illinois Shoreline. Illinois State Water Survey, Champaign, IL, Report of Investigation 110.
Corps of Engineers (COE). 1991. Precipitation on the
Great Lakes. Great Lakes Update 74:1–3.
———. 1993. The regulation of the outflow from Lake
Superior. Great Lakes Update 101:1–6.
———. 1997. High Water Level Concerns. //sparky.nce.
usace.army.mil/levels.
———. 2000. Water level controls. Great Lakes Update
139:1–5.
———. 2001. Current Great Lakes studies. Great Lakes
Update 145:4–5.
———. 2002. Are Great Lakes Water Levels Recovering?
Great Lakes Update 148:1–4.
Croley, T., Quinn, F., Kunkel, K., and Changnon, S.A.
1998. Great Lakes hydrology under a transposed climate. Climatic Change 38:405–433.
Day, P.C. 1926. Precipitation on the drainage area of the
Great Lakes. Monthly Weather Review 54:85–106.
Great Lakes Commission. 1986. Water Level Changes.
Ann Arbor, MI.
Hartmann, H. 1988. Potential Variations in Great Lakes
Water levels: A Hydrologic Response Analysis.
GLERL, Ann Arbor, MI. NOAA Technical Memorandum ERL 68.
Horton, R.E., and Grunsky, C. 1927. Hydrology of the
Great Lakes. Part 3, Appendix 2, Chapter 7. Report of
the Engineering Board of Review of the Sanitary District of Chicago, Chicago.
Horvath, F.J., Jannereth, M.R., and Shafer, C. 1988.
Impacts of Water Level Fluctuations. Land and Water
Management Div., Michigan Dept. of Nat. Resources,
Lansing, MI.
200
Stanley A. Changnon
Karl, T.R., and Knight, R.W. 1998. Secular trends in
precipitation amount, frequency and intensity in
the U.S.A., Bulletin Amer. Meteoro. Soc. 79:231–
241.
Lettenmaier, D.P., Wood, E., and Wallis, J. 1994.
Hydro-climatological trends in the continental U.S.,
1948–1988, Journal Climatol. 7:586–607.
Muller, F.B., Gervais, J., and Shaw, R. 1965. The Effect
of Basin Precipitation on the Level of Lake Michigan/Huron. Canada Department of Transport,
Toronto, Circular 4264 Meteorological Branch.
National Geographic. 2002. Down the drain? The incredible shrinking Great Lakes.
Quinn, F.H. 1978: Lake Superior regulation effects.
Water Resources Bulletin 14:1129–1140.
——— . 1999. Anthropogenic changes to Great Lakes
water levels. Great Lakes Update 136:1–4.
———, and Croley, T.E. 1981. The role of precipitation
climatology in the hydrologic design and planning on
the Laurentian Great Lakes. In Preprints 4th Confer-
ence on Hydrometeorology, pp. 7–11, American
Meteorological Society, Boston.
———, and Sellinger, C.E. 1990. Lake Michigan record
levels of 1838, A present perspective. J. Great Lakes
Res. 16:133–138.
———, Croley, T., Kunkel, K., and Changnon, S. 1997.
Laurentian Great Lakes hydrology and lake levels
under the transposed 1993 river flood climate. J.
Great Lakes Res. 23:317–327.
Water Science and Technology Board. 1989. Great
Lakes Water Levels: Shoreline Dilemmas. National
Research Council, National Academy Press, Washington, D.C.
Working Committee 2. 1993. Land Use and Management. Annex 2, Great Lakes-St. Lawrence River Basin
Levels Reference Study, Toronto, Canada.
Submitted: 28 May 2003
Accepted: 9 December 2003
Editorial handling: Barry M. Lesht