Introduction

Past studies have shown that freeze-up and ice-loss dates are useful indices of climate and climatic change (Reference Palecki and Barry.Palecki and Barry, 1986; Reference RobertsonRobertson, 1989; Reference SchindlerSchindler and others, 1990; Reference SkinnerSkinner, 1993). Long-term freeze-up and ice-loss records at Grand Traverse Bay and Lake Mendota began in the 1850s and by about 1890 the average dates reveal a shift to later freeze-up and earlier ice loss. A gradual warming during the 20th century was associated with a second shift to earlier ice-loss dates, starting in about 1940 at Grand Traverse Bay (1940-93) and in 1980 at Lake Mendota (Reference Assel and Robertson.Assel and Robertson, in press). Here we expand on this work by examining four additional sites spread throughout the Great Lakes region. The scope of this study is limited to a first exploratory look at the data. We examine long-term averages and variances (standard deviation) and smoothed plots of the annual data. Our objectives are to (a) identify similarities in the changes in ice cover over the past 200 years throughout the Great Lakes region, (b) describe differences in the ice cover among sites, and (c) qualitatively infer changes in the winter climate throughout the region over the past 200 years.

Data and Site Descriptions

Ice-event dates, i.e freeze-up and ice-loss (break-up) dates, are available throughout the Great Lakes region. Freeze-up can be defined as the date of the first skim ice, the date solid ice forms that lasts until the spring, or something in-between. Ice-loss date can be defined as the date the ice cover breaks up, the date of final loss of ice cover, or something in-between. Since a standard definition thus does not exist for freeze-up or ice loss, climatic analysis of these data is problematic. Albeit with this caveat, we found six sites, spanning virtually the entire region (Fig. 1), that had long-term freeze-up and ice-loss date data sets. These locations were (1) Chequamegon Bay (46.67° N, 90.88° W) between Honghton Point and Ashland, (2) Grand Traverse Bay (44.76° N, 85.62° W) from the head of the west arm of the bay to Marion Island, (3) Lake Mendota (43.66° N, 89.40° N), (4) Green Bay in the vicinity of Menominee, Michigan (45.00° N, 87.64° W), (5) Buffalo Harbor (42.88° N, 78.90° W) westward to Long Point, and (6) Toronto Harbor (43.65° N, 79.45° W) between Toronto Island and the mainland. The period of record and the average water depth for the six sites ranged from 64 to 144 year and from 7 to 46 m, respectively (Table 1).

Fig. 1.

Site location map.

Table 1.

Long-term average (Avg.) (in calendar days) and standard deviation (S.D.) of freeze-up and ice-loss dates.

Missing data in the time series were approximated in the following manner. Ice reports from the National Weather Service were used to estimate ice-loss dates for Buffalo Harbor and the east end of Lake Erie for the winters of 1946-50. Grand Traverse Bay freeze-up and ice-loss data (abstracted from Reference SniderSnider (1974)) provide estimated freeze-up dates for 1859, 1908, 1909, 1912, 1914, 1918, 1920 and 1921 and estimated ice-loss dates for 1852, 1906, 1908 and 1921. Also, there were 26 winters during which a solid-ice cover did not from in Grand Traverse Bay. The freeze-up and ice-loss dates for these winters were set to 8 March, the average date between the freeze-up and break-up dates for the five winters with the shortest ice durations.

Freeze-Up and Ice-Loss Dates

The long-term average freeze-up date is latest at Grand Traverse Bay (February “15.5”): 56.1, 57.4 and 58.4 d later than at Menominee, Lake Mendota and Toronto Harbor, respectively. The long-term average ice-loss date is latest at Buffalo Harbor (April “23.2”): 1.8, 17.8, 18.5, 22.7 and 24.7 d later than at Chequamegon Bay, Menominee, Lake Mendota, Grand Traverse Bay and Toronto Harbor, respectively. Air temperature is the single most important climatic variable affecting ice formation and loss, though water depth and water motion are also important site factors. Freeze-up is latest at Grand Traverse Bay because of its large heat storage (46 m average water depth) and mixing- with Lake Michigan. Ice loss is latest at Buffalo Harbor because of ice advection from Lake Erie.

The standard deviations of the freeze-up and ice-loss dates, taken together, are smallest at Lake Mendota (10.9 and 11.1 d respectively) and are likely to be similarly small at Chequamegon Bay (no data for freeze-up, 9.1 d for ice loss). Standard deviations in ascending order of magnitude after this are (1) Menominee (10.2 and 17.6 d), (2) Grand Traverse Bay, (17.2 and 17.9 d), (3) Buffalo Harbor (no data on freeze-up, 19.3 d for ice loss), and (4) Toronto Harbor (19.6 and 18.1 d). Lake Mendota, and Chequamegon Bay are relatively shallow, have minimal mixing from connecting rivers or adjacent water bodies, and thus have a relatively rapid decline in water temperature in autumn, resulting in the observed small variances in freeze-up and ice-loss dates. In the case of the Menominee site, which is also shallow and has a low variance in freeze-up dates, the timing and volume of increased Menominee River flow due to snow melt undoubtedly affect the date of ice loss.

Changes and Trends in Freeze-Up and Ice-Loss Dates, 1823-1994

To identify changes and trends in the ice data, a 10 year moving average was used to smooth the data. The smoothed data are plotted in the tenth year of the average.

Changes in Freeze-Up Date

The general trend shown in Figure 2a is toward later freeze-up dates, with must of the shift occurring before 1900. This is consistent with the end of the “Little Ice Age” around 1850 (Reference Lamb and LambLamb, 1982). Although only three of the locations have records predating 1900, all three indicate later freeze dates after the 1890s. The early data from Lake Mendota and Grand Traverse Bay apparently suggest two different climatic scenarios. that is, freeze-up dates from Grand Traverse Bay suggest a cooling trend with earlier freeze-up dates during the 1850s and early 1870s. while freeze-up dates from Lake Mendota appear to suggest just the opposite. However, the record at Toronto Harbor suggests the earlier ice dates at Grand Traverse Bay were probably just a short-term condition in an overall trend toward later freeze dates. Toronto Harbor’s freeze-up dates show a quasi-sinusoidal pattern during the entire period of record (1823-1919). The cause of the sinusoidal pattern is unknown and is currently under investigation.

Fig. 2.

10 year moving average (plotted on tenth year) of (a) freeze- up dates and (b) ice-loss dates.

Freeze-up dates since about 1900 have remained relatively stable. However, there was a greater frequency of winters with mild mid-winter temperatures from the early 1930s to the late 1950s (Reference AsselAssel, 1980) and again from the early 1980s to the early 1990s, resulting in an increased frequency of winters without freeze-up at Grand Traverse Bay which suggests trends toward later freeze-up dates in the 1930s, 1950s, and mid-1980s to early 1990s (Fig, 2a). Similar trends are absent from the time series for Lake Mendota and Menominee, likely because their freeze-up dates represent an earlier (November and December) air-temperature integration period. Grand Traverse Bay freeze-up dates represent a later (January) air-temperature integration period, closer to the ice-loss dates.

The changes in average freeze-up dates for Lake Mendota and Grand Traverse Bay occurring around 1890 were quantified in terms of changes in air temperature, using ice models by Reference Assel and Robertson.Assel and Robertson (in press). The result suggest an increase in early-winter air temperatures of approximately 1.5° C. However, freeze dates were in a period of transition, so air-temperature changes might be expected to be even larger if the period were extended back to the beginning of the Toronto record (1823).

Concluding Remarks

Air temperature is the primary climatic variable affecting freeze-up and ice-loss dates, while water depth and mixing with larger adjacent water bodies are also contributory factors affecting ice-event dates. Sites having greater water depth and more exposure to larger bodies of water have later freeze-up dates and greater variance in ice-event dates. The long-term average ice-loss date at such sites can be either about the same (Grand Traverse Bay) or much later (eastern end of Lake Erie) than the long-term average date at less exposed sites (Lake Mendota and Toronto Harbor).

The earliest ice records from Toronto, Grand Traverse Bay and Lake Mendota (1823-90) suggest that the winter climate near the Great Lakes prior to 1890 was warming as indicated by a trend toward later freeze-up and earlier ice-loss dates. This is in good agreement with the suggested timing of the end of the Little Ice Age, around 1850. Since about 1890, the average freeze-up dates have remained relatively steady, indicating little change in early-winter air temperatures. Ice-loss dates, however, have continued to become earlier as demonstrated by changes over relatively short periods (1) after 1940 at Grand Traverse Bay, (2) after 1980 at Lake Mendota. The reason for these changes appears to be that ice-event dates at Grand Traverse Bay represent an integration of air temperatures over different time periods and a different lake environment than at Lake Mendota. Ice records at all six sites indicate the warmest decade over the past 200 year occurred in the 1980s.

A key to quantifying the ice-event information in terms of climate change is to develop an objective and accurate ice-event dates model which relates ice events to air temperature or other climatic variable(s). We plan to explore the development of ice-event–air-temperature models to represent air-temperature integration periods associated with freeze-up and ice-loss dates at the four sites not examined by Reference Assel and Robertson.Assel and Robertson (in press). We also plan to examine possible teleconnections between ice-event dates in the different lake environments represented by the six ice sites and possibly at other sites within the Northern Hemisphere by comparing geopotential heights in upper-air data for the Northern Hemisphere.