Introduction

The Laurentian Great Lakes, shown in Figure 1, are one of the major water resources of North America. Systematic observations of the ice cover of the Great Lakes have been made for over 20 years by United States and Canadian government agencies in support of activities in the fields of research, operations, and planning. The ice cover is important because it has a major impact on the economy of the Great Lakes region through its effects on navigation, hydropower generation, lake levels, flooding, and damage to shore structures. Reports describing the annual ice cycle for individual winters have been published during the past 20 years and many of them are included in the selected bibliography for this paper.

Fig. 1. Map showing the relative positions of the Great Lakes and Lake St Clair (C).

An initial analysis of the ice-cover data was made by Reference RondyRondy (1969). He produced an ice atlas of normal ice cover on the Great Lakes for early winter. midwinter, and for the periods of maximum ice extent and early ice decay based on a subjective analysis of data from six winters. The Great Lakes-St Lawrence Seaway Winter Navigation Demonstration Program, authorized by the United States Congress in 1970 and completed in 1979, provided the major impetus to revise Rondy’s ice atlas because of the need for greater detail in ice-cover data to support increasing operational and research activities.

This paper describes a database containing information spanning a established for the rev highlights results of a respect to the extremes progression of ice cove detailed description of will be given in a fort 1 i shed by NOAA. The dat National Snow and Ice D Environmental Satellite (HESOIS) in Boulder, Co period of 20 years which was ision of Rondy’s atlas and statistical analysis with variability, and normal r on the Great Lakes. A the statistical analysis, hcoming ice atlas to be pub-abase will be archived at the ata Center of the National Data and Information Service orado, USA.

Description of the Database

Ice charts of the Great Lakes for the 20-year period from 1960 to 1979 have been produced by the organizations listed in Table I. The ice charts have appeared in several different map projections, at different scales, using different ice-cover symbols and terminologies. Over 2 800 historic ice charts were digitized for this study, reducing the abstracted data to a common polyconic map projection, as well as a common ice-concentration relating to ice concentration (the percentage of a unit surface area of the lake that is ice-covered) was abstracted from the original ice nearest 10% from a range of zero to 100% ice cover. An electronic digitizer coupled with used to abstract, code, and record the ice concentrations from individual ice charts The digital data were then converted to 5 × 5 km grid cells in a rectangular Cartesian coordinate system having apolyconic map projection. An example of the configuration of overwater grid cells for Lake Er Clair is given in Figure 2. The matrix dimensions for each Great Lake and the number of ized are given in Table II.

Table 1. Sources Of Historic Ice Charts Of The Great Lakes

Fig. 2. Configuration and location uf the 1041 over-water grid cells, indicated by dots (.) in the 40 × 91 digital data matrix of Lake Erie and Lake St Clair.

Table 2. Ice-Concentration Database For The Great Lakes (1960 to 1979).

Methods Of Analysis

The ice-concentration data set for each Great Lake was partitioned into half-month intervals for statistical analysis. This interval was chosen because it is the smallest averaging period for which there are sufficient data for analysis. Each half-month period was coded as follows: 16-31 December D2 1-15 January Jl 16-31 January J2 1-14 February Fl 15-28 February F2 1-15 March Ml 16-31 March M2 1-15 April Al 16-30 April A2 The number of charts for each half-month period is given in Table III. The time series of ice concentrations for each grid cell for a given period was analyzed for minimum and maximum ice concentrations. Subperiod ice concentrations for a given cell for which there were multiple observations for a given year were averaged; then the average, median, and mode ice concentrations were calculated for the resultant time series at each grid cell. A lakewide average value was then calculated for each half-month ice-concentration statistic by averaging all grid cells. The lakewide average is used to indicate the lakewide trend of each statistic over the nine half-monthly periods and for making comparisons with other lakes.

Table 3. Number Of Ice Charts Used For The Statistical Analysis Of Ice Concentration For Each Great Lake For Each Of The Nine Half-Month Periods Over The 20-Year Period 1960-79 (For Period Codes, See Above

Discussion Of Results

Lakewide averaged values of the maximum, minimum, average, median, and mode ice concentration are shown in Figure 3. The maximum and minimum curves illustrate the extremes in ice cover, while the median, mode, and average are used to estimate "normal" ice concentrations. Ice charts depicting patterns of ice-cover distribution for the maximum, minimum, and median ice concentrations over the nine half-monthly periods will be presented in the NOAA ice atlas scheduled for publication in 1983.

Fig. 3. Lakewide average value for each half-month ice-concentration statistic (maximum, minimum, average, mode, and median) for the nine half-month periods starting with the last half of December (D2) and continuing through the last half of April (A2) for each of the Great Lakes. For period codes, see page 15.

Ice-Cover Variability

The difference between lakewide averaged maximum and minimum ice concentrations is a measure of the variability of the ice co/er over the period of record for each half-monthly period. As shown on Table IV and portrayed in Figure 3, the ice-cover variability for the nine half-monthly periods is largest for Lake Erie, followed in order by Superior, Huron, Michigan, and Ontario. The ice-cover variability for each half-monthly period averaged over the five Great Lakes is also given in Table IV. As might be expected, ice-cover variability is smallest during the period of initial ice formation and the period of final ice loss, when it averages approximately 35% for the five lakes. At its maximum, during the last half of February, ice-cover variability averages 38%. The latter half of February is the half-month when ice cover is normally at its greatest area extent, so that the greatest difference in ice cover over the 20 winters should logically occur during this half-month period. Ice-cover variability over the nine half-monthly periods for individual lakes is indirectly related to mean lake depth, which is the ratio of the volume to the surface area, and is directly related to cumulative freezing degree-days (FDDs), which is a measure of the cumulative departure of the mean daily air temperature from 0°C over the winter season. Mean values of ice-cover variation, lake depth, and FDDs are given in Table V.

Table 4. Variation Of Ice Concentration (Lakewide Average Of Half-Monthly Maximum Minus Lakewide Average Of Half-Monthly Minimum) Expressed As A Percentage For Each Lake For The 20-Year Period From 1960 To 1979 (For Period Codes, See Page 15)

Table 5. Mean Lake Depth, Freezing Degree-Days (Fdds) And Averaged Ice-Cover Variability

Lake Erie and Lake Huron, with mean depths of 19 and 59 m respectively, have less heat to lose before ice forms than have the other Great Lakes and so can develop more extensive ice covers early in the season than can the other Great Lakes. Thus, in early winter the ice-cover variation is greater for these two lakes compared to the other lakes. Ice-cover variabil-ranges from 64% on Lake Erie to 17% on Lake Ontario during the second half of December.

Ice-cover variability for February and the first half of March averages over 80% for the Great Lakes and is related to FDD accumulation; for more detail, see studies by Reference RichardsRichards (1964) and Reference RogersRogers (1976). In most years, above-freezing air temperatures (i.e. negative FDDs) in early March bring a reduction in ice cover. This is reflected in decreasing ice-cover variability, first on Lake Ontario during the first half of March, next on Lake Michigan during the second half of March, and finally on the remaining lakes during April. In the case of Lake Erie and Lake Huron, high ice-cover variability occurs up to mid-April, while in the case of Lake Superior, and the northern sections of Lake Michigan and Lake Huron low air temperatures can preserve existing ice covers into May during some years, resulting in high ice-cover variability through the second half of April.

Normal Ice Cover

In general, ice covers are restricted to the bays, harbors, and constricted areas of the Great Lakes through the end of January, with the exception of Lake Erie, where ice covers form in mid-lake in January owing to the shallow water. Ice forms in the mid-lake areas of the Great Lakes in February and ice cover reaches its maximum areal extent in this month or early March. During the time of maximum ice extent, the median ice concentration on the Great Lakes is: Erie 90%, Superior 75%, Huron 68%, Michigan 45%, and Ontario 24%.

Ice covers begin to decline in extent in March. The largest decrease in ice cover for Lakes Michigan and Ontario occurs during the first half of March. The largest decrease in ice cover for Lake Erie occurs during the second half of March, although Lake Erie also has a significant decrease in ice cover during the first half of March. The greatest loss in ice cover for Lakes Superior and Huron occurs during the first half of April. By the second half of April, less than 10% of the surface area of the Great Lakes is normally 1ce-covered. The remaining ice is usually located in bays, harbors, along windward shores, and in constricted areas.

Concluding Remarks

A 20-year computerized ice-concentration database has been established for the Great Lakes and used for a statistical analysis of the ice cover for nine half-monthly periods. Lakewide trends for maximum, minimum, median, mode, and average ice-concentration statistics over the nine half-monthly periods have been identified. Ice-cover variation is defined by the difference between lakewide average maximum and minimum ice concentrations for individual lakes. The median, mode, and average statistics were used to estimate the normal extent of ice cover.

Ice charts depicting the maximum, minimum, and median ice cover for the nine half-month periods will be published as a NOAA ice atlas. The entire digital ice-concentration data set will be transferred to the National Snow and Ice Data Center, HESDIS, Boulder, Colorado, DSA and documented by a report describing its format and structure. It is hoped that these data will prove useful to a broad spectrum of users in the fields of research, operations, and planning. Further analysis of the data will include spatial and temporal correlations of the ice concentration, as well as examination of the relationship of ice concentration with lake depth and air temperatures on sub-lake spatial scales.