Lake ice climate normals for the Great Lakes 1991 to 2020

Preface

The Canadian Ice Service (CIS) Regional Ice Chart database (CIS Ice Archive) encompasses over 50 years of ice information for the:

  • Western Arctic (1968 to present)
  • Eastern Arctic (1968 to present)
  • Hudson Bay (1971 to present)
  • East Coast (1968 to present)
  • Great Lakes (1972 to present)

In these Canadian Ice Service lake ice climate normals, weekly regional charts depicting lake ice concentration and stage of development were used to compute a set of 30-year climatological standard normals for the ice seasons spanning November 1990 to 1991 through June 2019 to 2020.  Lake ice in the Great Lakes typically begins to form in November of one year and melts away in the spring of the following year. This publication for the Great Lakes follows from two previous 30-year climate normals for the periods 1973 to 2002 and 1981 to 2010, where the indicated years represent the end-years of the winter ice seasons.

CIS 30-year climate normals are prepared in accordance with World Meteorological Organization (WMO) guidelines. The lake ice climate maps herein provide a week-by-week depiction of normal ice conditions throughout each winter: its location, distribution, extent, thickness, and variability. They are used as the baseline for computing departure-from-normal ice concentration products and to provide guidance for planning safe operations in ice-infested Canadian waters.

Noteworthy in this edition

Chart data reliability: These 1991 to 2020 climate normals are the first 30-year CIS normals in the series where the majority (approximately 80%) of the chart data used was produced: a) digitally, using Geographic Information System (GIS) software that was introduced in 1995, and b) using primarily high-resolution (approximately 50m to 100m) Synthetic Aperture Radar (SAR) satellite imagery with the launch of Radarsat-1 in 1995. Producing ice charts using GIS software has reduced the impact of errors associated with the digitization of paper charts, while using SAR imagery has significantly reduced analysis-related uncertainties (e.g. due to cloud contamination) and greatly improved the identification of old ice.

These climate normals are available on the Canadian ice Service website

Recommended citation: Canadian Ice Service. 2021. Lake ice climate normals for the Great Lakes 1991 to 2020.

1. Canadian Ice Service regional ice charts: production method and quality control

1.1 Chart preparation: satellite imagery and additional observations used

The data used in the production of these climate normals are derived from Canadian Ice Service (CIS) weekly regional ice charts (Figures 1.1 and 1.2), which are created using a combination of remote-sensing data and observations made by Ice Service Specialists (ISS) onboard dedicated aircraft and Canadian Coast Guard (CCG) ships.

Between 1991 and 1996 the proportion of Canadian waters covered by satellite data available for the production of regional ice charts increased from approximately 50% to approximately 80% with the launch of the Canadian Space Agency’s (CSA) RADARSAT-1 satellite. Pre-1996, aerial surveillance (both visual and using Side-Looking Airborne Radar – SLAR) was the dominant data source along with optical imagery from United States Geological Survey’s (USGS) Landsat-4, synthetic aperture radar (SAR) data from the European Space Agency’s (ESA) ERS-1, visible and infrared data from the National Oceanic and Atmospheric Administration’s (NOAA) AVHRR and passive microwave data from the National Aeronautics and Space Administration’s (NASA) SSM/I sensors. From 1996 onward, high resolution SAR data available from the CSA’s RADARSAT missions quickly became the dominant source of satellite imagery used in ice chart preparation. The current normals are the first 30-year CIS climate normals to be produced from predominantly SAR-based charts (greater than 80%).

Since 1996, the main source of SAR imagery, typically 50 m to 100 m resolution, is from CSA’s RADARSAT-1 (1996 to 2013), RADARSAT-2 (2008 to 2020) and RADARSAT Constellation Mission (2020-present).  These data have been supplemented with various sources of optical, passive microwave and additional SAR imagery over the years.  The primary sources of additional SAR imagery are the European Space Agency’s (ESA) ERS-1 and ERS-2 (1991 to 2010), Envisat (2002 to 2012) and Sentinel-1A and 1B (2015 to present) satellites. The primary sources of optical imagery used at CIS (visible, near-infrared and infrared), typically 250 m to 1 km resolution, are NOAA’s AVHRR (1981 to present), NASA’s MODIS (2000 to present), NOAA’s VIIRS (2012 to present) and most recently NOAA’s GOES-16 (2017 to present).  Passive microwave imagery is only used when other imagery is not available, resolution ranges from 12.5 km to 25 km, and data is available from NASA’s SSM/I-SSMIS (1987 to present) and AMSR-E (2002 to 2011), and the Japan Aerospace Exploration Agency’s (JAXA) AMSR2 (2012 to present).

The main advantage of SAR data is that it is independent of solar illumination and cloud cover. Under cold conditions, SAR imagery also provides a very clear distinction between seasonal ice and multi-year ice, where multi-year ice appears brighter due to a much larger concentration of air bubbles, lower salinity and generally rougher ice surface. These two factors alone eliminate large sources of uncertainty in the production of CIS ice charts.

SAR imagery does have some disadvantages, and in these instances optical imagery and sometimes passive microwave imagery can provide valuable information when producing ice charts. Compared to optical imagery, SAR imagery has a much smaller swath width (hundreds versus thousands of kilometers) and longer revisit frequency (days versus multiple times a day). In ice charting, it is difficult to distinguish between very smooth ice and calm open water in SAR imagery.  Optical imagery provides a very clear distinction between ice and water, making it useful for detecting and verifying open water polynyas and flaw leads in ice packs. During the summer months in the Arctic in the presence of wet snow and when melt water floods the surface of ice floes, there is a loss of definition between first year ice and old ice in SAR imagery, and sometimes between open water and the ice floes themselves.  Again, optical imagery does not suffer the same loss of detail and is a useful supplement to SAR data.

1.2 Chart preparation: methodology and quality assurance

1.2.1 Methodology

In 1996, regional chart production at CIS transitioned to an entirely digital workflow with the introduction of the Ice Service Integrated System (ISIS). It was built on Leica Geosystems’ Erdas Imagine and ESRI’s ArcGIS ArcInfo, which was still in use in August 2021. Prior to 1989, the personnel at CIS drew by hand the regional charts on paper. Between 1989 and 1995, source data was displayed digitally, on a common projection and scale, in the Ice Data Integration and Analysis System (IDIAS).  Although digital chart production was possible during this time, the system was slow and was primarily used as a tool to view data.  For climatological purposes, all hand-drawn charts were digitized in the late 1990’s and added to the digital archiveFootnote 1 . It should be noted that the original paper map scale of the Regional Ice Charts was 1:4,000,000.  The amount of detail and accuracy in the current GIS-produced analyses is comparable to the original maps. The 1991 to 2020 climate normals are the first 30-year CIS normals based predominantly on digitally produced ice charts (greater than or equal to 80%) significantly reducing the error associated with digitizing paper charts and sourcedata that is not displayed on a common scale or projection for chart production.

For chart production beginning in 1996, satellite imagery and daily ice charts are imported into GIS software (ISIS) and analyzed into the World Meteorological Organization’s (WMO) standardized egg code, which contains information for ice concentration, ice type, and floe size.  To identify ice type (an indication of age and thickness) in satellite imagery, various indicators such as brightness, fracture patterns and the shape of ice floes are used. The experience of the analyst plays an important role in chart production, meteorological and oceanographic factors such as air temperature, winds, water temperature, salinity, currents, waves, and tides, along with climatology to supplement information from satellite imagery and daily ice charts are considered.   In addition, ground-truth data provided by ISS’s on CCG ships and aircraft are incorporated when available. A notable difference between the Great Lakes, East Coast and Arctic regional ice charts is the reliance on daily ice charts during production.   The Great Lakes and East Coast regional charts are generated primarily from daily ice charts whereas the Arctic regional charts are generated primarily from satellite imagery with the daily ice charts used as guidance.

1.2.2 Quality assurance

Quality Assurance steps taken during the preparation of regional ice charts are both automated and manual.  Automated checks are from GIS scripts that scan for: a) errors when coding partial concentrations and floe sizes, and b) inconsistencies in overlap areas between charts (Figure 1.1) for adjacent regions.  Manual checks during production ensure that the analysis is consistent with: a) the climatology for a given region and time of year, and b) with the expected ice thickness associated with the accumulated freezing degree days (FDDs) during the freeze-up and winter seasons.  After production, the team leader or another dedicated person in operations manually inspects each ice chart for errors and inconsistencies before it is disseminated, archived and published online.  Chart errors noticed internally in operations at CIS after a chart has already been issued are corrected and a chart labeled “amendment” or “correction” is disseminated, archived and published as soon as possible.

A map of the East Coast of Canada depicting coloured areas of ice concentration and descriptions of ice type present on March 3, 2014.
Figure 1.1. Canadian Ice Service regional ice chart extents. The climate normals for Northern Canadian Waters is derived from charts for the Western Arctic, Eastern Arctic and Hudson Bay. Where northern charts overlap, although consistency between charts is ensured during the chart production stage, the data from the Eastern Arctic chart only was used. Note that with the exception of the Great Lakes, all regional chart extents changed in 1997 with the switch from hand-drawn to GIS charts.
A map of the Great Lakes depicting coloured areas of ice concentration and descriptions of ice type present on March 10, 2014.
Figure 1.2. Example weekly regional ice chart (Great Lakes)

1.3 Regional chart archive: quality control and homogenization

1.3.1 Chart dates: original chart dates vs “Historical Dates”

1.3.1.1 Original dates

Regional Ice Charts are currently produced weekly throughout the entire winter ice season for the Great Lakes.

Although the date specified on the charts is the Monday of each week, they represent a rough approximation of ice conditions for a seven-day period centred on this day and are finalized and issued on the Wednesday of that week. Satellite imagery up to 2 to 3 days on either side of the Monday may be consulted during the production of a regional chart but preference is given to images 2 to 3 days before to 1 day after the Monday. 

1.3.1.2 Historical Dates

Since the Monday of any given week does not fall on the same date each year, the charts in the database are assigned a “Historical Date” for consistency between years. The Historical Dates (HDs), one per week, are every 7 days starting January 1.  The chart date is assigned to the closest Historical Date.

There are two exceptions to this 7 day progression in HDs:

  1.  the 8-day period between February 26 and March 5 that occurs in leap-years
  2.  the 8-day period between November 26 and December 4 that accounts for the fact that there are 365 days in a year, not 364.

The week of November 26 is chosen to have 8 days because it is a week after the shipping season in northern waters has ended and before the ice forming in southern waters has an impact on shipping.

1.3.2 Missing charts

The World Meteorological Organization (WMO) recommends that:

  • statistical normals be computed when data is available for at least 80 percent of the years in a 30-year averaging period
  • where data is missing, it should not be missing for more than three consecutive years (WMO, 2011: Guide to Climatological Practices (WMO-No.100), Geneva).

CIS weekly regional ice charts are occasionally missing in the archive if:

  1. a paper chart from the hand-drawn era was not digitized
  2. a chart’s data was not sent to the archive
  3. it was not produced.

For the current 30-year period, all missing charts related to the first two cases have been located and re-digitized, and re-integrated into the database. Missing charts related to the third case were replaced using analogue charts rather than interpolation, as explained below. 

Charts that are missing for a HD because they were not produced often occur at the start or end of the ice season if chart production for that winter began after ice formation already started, or ended before all the ice completely melted away. This practice was more common prior to the availability of high-resolution SAR satellite imagery, and mainly impacts the statistics for certain bays in the Great Lakes and coastal areas of the Labrador Coast. Where charts are missing at the beginning or end of an ice season, interpolation of the data between adjacent weeks is not an option.  Instead, analogue replacements were chosen by reviewing the start and end of the ice season in question and any missing charts were replaced with charts from years with analogous patterns of ice formation and melt.

1.3.3 Chart coding / polygon errors

Occasionally, in spite of the Quality Assurance steps taken during the production of the Regional Ice Charts, errors in the coding of the ice do occur (e.g. very thick lake ice “1.” might erroneously be coded as new lake ice “1”).  In addition, during the digitization process of the pre 1996-97 paper charts, mistakes were made when transcribing the ice codes and in the interpretation of polygon boundaries.   

Beginning with the production of the 1973 to 2002 normals, a continued effort has been made over the years to manually inspect and correct as many archived charts as possible. Additional errors identified using automated scripts (e.g. where sea ice instead of lake ice codes were mistakenly used) were corrected during the production of this 1991 to 2020 climate normals.  While the majority of errors in the Regional Ice Chart digital archive have now been corrected, occasional errors may remain.

1.3.4 Temporal homogeneity

In addition to changes in data sources and chart preparation technology over the years (e.g. changes from optical to SAR satellite imagery, changes from hand-drawn to GIS based chart production), some changes in ice analysis practices have also occurred. For example, in 1982, the code used to describe the ice in each polygon changed from a “ratio” code to an “egg” code (for a description of the egg code, see chapter 3 in the Manual of Ice (MANICE). Around the same time, the encoding of first year sea ice changed from a single category to three sub-categories (thin, medium and thick).

Changes in ice analysis practices that impact the current normals period are outlined below, as well as the steps taken to address these discontinuities in the chart data to mitigate their impacts on the climate normals products.

  • 1996-1998: With the adoption of GIS software, the practice of allowing multiple ice eggs within a single large polygon was changed to allow only a single ice egg per polygon. To compensate for the earlier practice, polygons with multiple ice codes were divided into separate polygons during the digitization of the hand-drawn charts.
  • 2004-2005: Polygons containing fast ice can be encoded using either an ice egg or a point symbol.  Prior to 2005, fast-ice point symbols were generic, only indicating that the ice had a concentration of 10/10 and was fast.  As of 2005, however, fast ice point symbols were affixed with the stage of development (SOD) of the predominant ice type.  Because this change in practice falls in the middle of the 30-year data period of the 1991 to 2020 lake ice climate normals, steps were taken to address this discontinuity (as done for the 1981 to 2010 climate normals). Pre-2005 generic point fast ice values were assigned SODs based on accumulated Freezing Degree Days and available climatology of shore-based ice thickness observations. Note that although this method has verified well where seasonal ice is the predominant fast ice type, it can lead to occasional errors in the fiords of far northern Ellesmere Island where the fast ice is predominantly multi-year ice (i.e. where fast ice persists in fiords from one year to the next).  For this reason (and others – see the Eastern Arctic bullet in section 1.3.5 below) extreme high Arctic areas have been excluded from the climate normals.
  • 2014-2015: Prior to 2014-2015, fast ice was almost always considered to have a concentration of 10/10 (i.e. to be “consolidated”). Similarly, non-fast or mobile ice packs, with the exception of vast or giant ice floes, was almost always assumed to contain fractures and to therefore have a concentration of <10/10. Improvements in satellite image resolution (both optical and SAR), has now allowed for the definitions of fast and consolidated ice to be separated on a more frequent basis. As such, fast ice is now often analyzed as “fractured” during the melt season and non-fast ice is sometimes analyzed as “consolidated” during the winter where mobile ice has temporarily been compressed into a solid pack. This recent change in practice has an impact on the interpretation of the median ice concentration products in these normals.  Prior to this change in practice, areas of 10/10 median concentration served as a proxy for coastal fast ice extents throughout the year, something of great interest for community and marine navigation planning. With the new practice, climatological information regarding fast ice extents is no longer implicit within the median ice concentration maps. To address this discontinuity and to be consistent with past climate normals, all fast ice was assigned a concentration of 10/10 and all mobile, non-fast ice, was assigned a concentration of < 10/10 prior to computing median concentrations. Almost all changes were made to charts after 2013 (<25% of charts). To further address this issue moving into the future, new “fast ice”-specific map products not based on median concentrations (“Frequency of Presence of Fast Ice” and “Dates of Fast Ice Freeze-up and Breakup”) have been developed for these climate normals.

1.3.5 Spatial homogeneity

In addition to the technological- and analysis-related changes described above, changes in CIS ice chart coastlines have also taken place over the years, due to:

  1. a change in datum from NAD27 to WGS84
  2. improved geographic detail in some coastal areas.

Because such coastal variations are small compared to the scale of analysis in the regional ice charts, to compensate for spatio-temporal changes in coastline details and extents when producing the climate normals maps, a “Climate normals-coastline” comprising of the union of all coastline extents through time was used.  This decision has the potential to impact the climate normals in very small bays and inlets, but given the scale of the climate normals maps these minor small-scale issues were deemed insignificant and acceptable.

Note that while changes in CIS ice chart extents have taken place for the East Coast and Arctic regions over the years, the Great Lakes region has not been impacted by such a change (see Figure 1.1).

2. Description of climate normals statistics based on CIS ice charts

All statistics used in the climate normals are described below. Mathematical definitions of each statistic can be found in Appendix 1 (coming soon). Illustrations in terms of Ice Eggs can be found in Appendix 2 (coming soon).

2.1 Statistics by historical date

As with previous CIS normals, two key statistics have been selected to describe the 30-year climatological ice conditions in Canadian waters: medians and frequencies. These statistics are computed for various ice chart elements for each week (or Historical Date) in a given region’s ice season or year. For a complete understanding of the climatological ice conditions in any given location and time of year, median and frequency climate normals products should be used together.

Typical Ice Conditions

  • For information about “normal or average ice conditions” use the “Median of Ice Concentration” and “Median of the Predominant Ice Type**” products as a pair
  • For information about “normal or average old ice” (includes multi-year ice and second-year ice) conditions use the “Median of Old Ice Concentration” product
  • For information about the “normal or average extent of fast” (ice including the predominant locations of ice arches) use the “Median of Ice Concentration” product where an ice concentration equal to 10/10 is a proxy for fast ice.

Range of Ice Conditions

For information describing the range of ice conditions that could be encountered use the frequency of presence products.

  • For information about the “range of ice extent” (ice concentration greater than 1/10) including the “maximum and minimum ice extent” use the “Frequency of Presence of Sea Ice” product.
  • For information about “expected ice conditions considering only those years where ice is present”use the “Frequency of Presence of Sea Ice” product paired with the “Median of Ice Concentration When Ice Is Present” and the “Predominant Ice Type When Ice is Present“ products together.
  • For information about the “range of old ice extent” use the “Frequency of Presence of Old Ice” (old ice concentration greater than 1/10) and the “Frequency of Presence of Old Ice 4/10 to 10/10” products.
  • For information about the “range of fast ice extent” use the “Frequency of Presence of Fast Ice” product.

Typical timing of break-up and freeze-up

  • For information about the “typical timing of spring break-up and fall freeze-up” when the ice concentration shrinks below 1/10 or exceeds 1/10 use the “Date of First Ice” and “Date of Last Ice” products.
  • For information about the “typical timing of spring break-up and fall freeze-up” when ice concentration shrinks below 4/10 or exceeds 4/10 use the “Break-Up Date” and “Freeze-Up Date” products

2.1.1 Medians

Medians are used rather than averages because of the often abrupt day-to-day changes observed in sea ice concentrations (e.g. as witnessed during the fracture and break-up of fast ice areas during the melt season, or when wind-driven ice compression events are followed by abrupt ice dispersals when a weather system passes and winds change direction). 

To illustrate, consider a 5-week sample of ice concentrations in a given area during the break-up period with the following values: 10/10, 10/10, 10/10, 0/10, 0/10.  The average ice concentration for this period is (10 + 10 + 10 + 0 + 0)/5 equals 6/10. However, open drift (4/10 to 6/10) ice concentrations were never actually observed during this period. Although the ice may have briefly experienced open drift concentrations during its sudden fracture and dispersal, as a climatological normal it is not a useful measure. On the other hand, the median ice concentration for this period (determined by ordering the concentrations from lowest to highest and choosing the middle value) is 10/10, a realistic “normal” situation. The median is thus considered to be a more appropriate representation or measure of the true “normal” ice conditions over the 30-year climate normals period. Note that for even sets of values, as occurs with 30-year periods, CIS uses the higher, not the average, of the two middle values when computing medians. This is done to avoid fractional values and to err on the side of heavier ice conditions for marine safety.

2.1.1.1 Median of ice concentration

These products depict the 30-year median of the “total ice” concentration at any map location for each week of the ice season or year.  In these climate normals, areas of 10/10 median ice concentration can be used as proxies for fast ice extents on any given Historical Date within the 30-year period (see section 1.3.4 for caveats). This proxy should be used in conjunction with the new Frequency of Presence of Fast Ice products (described in section 2.1.2.3 below) for a clearer overall picture of normal fast ice conditions for specific weeks of the year.

2.1.1.2 Median of predominant ice type

These products depict the predominant ice type that is normally encountered at any map location for   each week of the ice season or year. Open water is considered an ice type in the calculation of the median predominant ice type.  This is a new product for the 1991 to 2020 climate normals.

“Predominant” is defined as the ice type with the greatest partial concentration except in cases where:

  • Partial ice concentrations are equivalent.
    If three ice types are reported and they all have equal partial concentrations, then the thickest ice type is deemed to be the predominant ice type (for marine safety reasons). If only two ice types are reported and they have equivalent concentrations, then again the thicker ice type is deemed to be the predominant ice type. Similarly, if more than two ice types are reported and the two thickest ice types have equal concentrations, then the thicker of the two is deemed to be the predominant ice type.

Once the predominant ice type at a given location has been determined for each year for a given Historical Date, the “Median of the Predominant Ice Type” is then computed by ordering the predominant ice types from thinnest to thickest and selecting their middle value (see Appendices 1 and 2).

2.1.1.3 Median of ice concentration when ice is present

These products depict the median of the total ice concentration but at any given point on the map, only use years when ice was present with a concentration ≥ 1/10 in the median calculations.  If in any one of the years only trace amounts or no ice was present at a certain location for the week (or Historical Date) in question, that year’s concentration of < 1/10 is omitted in the calculation of the median for that location and week. If in every one of the years no ice with a concentration ≥ 1/10 was ever present at that location and week, a median value of 0/10 is assigned to that location.

Interpretation- The most appropriate way to interpret these charts is to view them in conjunction with the 30-year frequency of presence of sea ice charts. To illustrate, suppose that for a given Historical Date a particular point has a  frequency of presence of sea ice with a range of 34% to 50% and a median of ice concentration when ice is present of 9/10 to 9+/10. This means that, for this Historical Date, this location has a 34% to 50% chance of experiencing sea ice and, when ice is present, it is “normally” 9/10 to 9+/10 concentration.

Important - Unlike in the “Median of Ice Concentration” products, areas of 10/10 median ice concentration in the “Median of Ice Concentration when Ice is Present” products do not serve as a reliable proxy for fast ice extents, particularly in the vicinity of fast ice margins during the spring to summer breakup period.  Because of the nature of the calculation, at these times and in these locations areas of 10/10 median ice concentration “when ice is present” are prone to appearing discontinuous and fractured, not a realistic “normal” situation. 

2.1.1.4 Median of predominant ice type when ice is present

These products depict the predominant ice type that is normally encountered when ice is present at any map location for a given week of the ice season or year.

“Predominant” is defined in section 2.1.1.2.

Once the predominant ice type at a given location has been determined for each year for a given Historical Date, the “Median of the Predominant Ice Type when Ice is Present” is then computed by ordering the predominant ice types from thinnest to thickest and selecting their middle value (see Appendices 1 and 2).

Interpretation: As with the “Median of Ice Concentration when Ice is Present” normals map products, the most appropriate way to interpret these charts is to view them in conjunction with the frequency of presence of sea ice charts. For example if, at a particular point, the 30-year frequency of presence of sea ice is in the range of 34% to 50% and the median of predominant ice type when ice is present is thin first-year ice, then at this point there is a 34% to 50% chance of encountering sea ice, and when ice is present, it is “normally” thin first-year ice.

2.1.2 Frequencies

Frequencies (expressed as percentages) are determined by summing the number of observations of an occurrence or event (e.g., the presence of sea ice during a specific week of the year) over the 1991 to 2020 climate normals period and then dividing by thirty, the total number of years in the period (see Appendices 1 and 2). 

These products can be interpreted as the probability of sea ice, or sea ice of a specific type and concentration, being present at a given location on a given Historical Date, based on the normal for the 1991 to 2020 period. The charts can also be used as proxies for maximum and minimum ice extents. The 0% line on the charts represents the maximum ice extent, beyond which no ice of the specified type and concentration was reported during the 30-year period for that Historical Date. The 100% line represents the minimum ice extent, within which there has always been ice of that type and concentration reported during the 30-year period for that Historical Date. In between these two lines, areas with frequencies of 1% to 33% represent above normal extents. Areas with frequencies of 34% to 66% represent near normal extents. Areas with frequencies of 67% to 99% represent below normal extents and can also be interpreted as those regions where the specified ice type and concentration is consistently found even during years with below normal ice extents.

2.1.2.1 Frequency of presence of sea ice (%)

These products depict the likelihood of encountering total ice concentrations greater than or equal to 1/10 at any particular location on any Historical Date during the 30-year period. For example if, at a particular point on the chart for a given Historical Date, the 30-year frequency of presence of sea ice falls within the category of 34% to 50%, then for this week of year and at this location there is a 34% to 50% chance of encountering sea ice with a concentration greater than or equal to 1/10. Furthermore, for this Historical Date, any ice encountered at this location lies near the normal limit of ice with total concentrations  greater than or equal to 1/10.

2.1.2.2 Frequency of presence of fast ice

These 30-year products are a new addition to the 1991 to 2020 climate normals. They depict the likelihood (on a weekly basis) of the ice being “fast ice” at any given point during the ice season or year.  For example if, at a particular point on a given Historical Date, the 30-year frequency of presence of fast ice is in the range of 67% to 84%, then at this time of year and at this location there is a 67% to 84% chance of encountering ice that is “fast”. Furthermore, for this Historical Date, any ice encountered at this location lies well inside the normal 30-year fast ice limit.

2.2 Dates of freeze-up and break-up

2.2.1 Dates of freeze-up and break-up – all ice

The “Date of First Ice, “Date of Last Ice”, "Freeze-up Dates” and “Break-up Dates" maps (derived from the median of ice concentration products) summarize the evolution of the ice extent on a biweekly basis during the freeze-up and break-up periods. 

To create these summary products, the median of ice concentration for every second Historical Date within the freeze-up period spanning December 11 to February 26 and the breakup period spanning March 5 to April 30 is consulted. Outside of the permanent Arctic ice pack, the date of first ice and freeze-up at any location is defined as the first occurrence of a median of ice concentration of 1/10 or greater and 4/10 or greater respectively, within the above set of freeze-up dates. The date of last ice at any location is defined as the last occurrence of a median of 1/10 or greater, within the above set up break-up dates. The date of break-up at any location is defined as the first occurrence of a median ice concentration less than 4/10, within the above set of break-up dates. 

Note that for the Great Lakes, in some years new ice may begin to form in late November (prior to the start of the freeze-up period considered above), and in some years ice will linger in the Thunder Bay area of Lake Superior until May (past the last breakup date defined above).  However these represent extreme years and are not reflected once median concentrations are computed.

2.2.2 Dates of freeze-up and break-up – fast ice

The "Fast Ice Freeze-up Dates” and “Fast Ice Break-up Dates" maps summarize the changes in the fast ice extent on a biweekly basis during the freeze-up and break-up seasons. Unlike the freeze-up and break-up dates computed for all ice, this calculation is based on the frequency of presence of fast ice and not the median ice concentration (because median ice concentrations of 10/10 are only a proxy for fast ice extents – see section 2.1.1.1). Fast ice freeze-up dates are defined as the first week the frequency of fast ice at a given location is  greater than or equal to 50%, while breakup dates are defined as the first week the frequency of fast ice is less than 50%.

3. Regional ice regimes and influences

3.1 Regional notes

A map of the Great Lakes with water currents and bathymetry
Figure 3.1 Great Lakes bathymetry and water currents

Regional Ice facts – Great Lakes (PDF)

Factors influencing Great Lakes ice extent and thickness

Air temperatures

Ice forms and thickens when air temperatures drop below freezing and water temperatures cool. While winter air temperatures can fluctuate greatly due to winter storms, the northern lakes are more likely to experience consistently cold temperatures through the winter months. For this reason, northern areas are more likely to see thicker ice at the end of the season, and the last ice to melt is typically found in the northern bays of Lake Superior and in the North Channel of Lake Huron.

Water depth

Because shallow waters cool more quickly than deeper waters, ice forms first in coastal areas and this is usually where the thickest ice will be found. Lake Erie, being very shallow, is often completely ice covered in mid-winter. Other lakes with deeper basins will often remain ice free in their central parts.

Winds and storms

Winds and transient winter storms modify the distribution and form of the ice. Warm air from the south can melt thinner areas of ice. Large waves can break up areas of ice into smaller floes. Where winds push the ice away from the shore, it will disperse and open water leads may form.  Where winds push the ice against a shore, it will become compact and may pile up into ridges.

Median Ice Season is defined as the period during which median ice concentrations in the region are ≥1/10.

Lake Superior

Median ice season: late-November to mid-May

Latest ice presence: early-June

Max ice cover: mid-March (37%)

Variability in max cover: 10% - 98%

Ice thickness: 45-85cm (along coast)

Max thickness: ~25m (pressure ridges)

Special ice features: In Whitefish Bay, shallow water and funneling of winds from the northwest cause ridging and compressed lake ice during the winter. Ice tends to be thicker than the rest of the mobile ice in Lake Superior.

Lake Michigan and Green Bay

Median ice season: early-December to mid-April

Latest ice presence: early-May

Max ice cover: mid-February (20%)  [Green Bay 100%]

Variability in max cover: 12% - 88%

Ice thickness: 45-75cm (coastal harbours and bays) 

Max thickness: 25-35m (ridges in Straits of Mackinac)

Special ice features: Due to prevailing westerly winds, the Straits of Mackinac often becomes a congestion point for lake ice, with ice often converging and ridging, leading to some of the thickest ice on the Great Lakes.

Lake Huron and Georgian Bay

Median ice season: early-December to late-April

Latest ice presence: mid-May

Max ice cover: mid-February (43%)  [Georgian Bay late-February 85%]

Variability in max cover: 25% - 98%

Ice thickness: 45-75cm (along coast)

Max thickness: up to 18m (pressure ridges)

Special ice features: The lake ice in the St. Mary’s river and North Channel becomes land fast during the winter. Breakup of the ice in these areas in the spring is influenced by icebreaking operations, speeding up the ice melt and breakup process.

Lake Erie and Lake St. Clair

Median ice season: mid-December to mid-April

Latest ice presence: early-May

Max ice cover: mid-February (85%) [Lake St Clair 100%]

Variability in max cover: 8% - 100%

Ice thickness: 25-45cm (in coastal bays)

Max thickness: up to 20 m (pressure ridges)

Special ice features: Because of its shallow profile the water temperature can change fairly rapidly.  This causes lake ice to form and melt much more rapidly in this lake than the other Great Lakes. Atmospheric temperature changes between above and below freezing can cause large fluctuations in ice cover on the Lake.

Lake Ontario

Median ice season: end-December to early-April

Latest ice presence: late-April

Max ice cover: mid-February (14%)

Variability in max cover: ≤10% - 65%

Ice thickness: 20-60cm (in bays)

Max thickness: significantly >60cm (pressure ridges)

Special ice features: Lake Ontario has the lowest maximum ice coverage of all of the Great Lakes owing to its depth and its relatively warmer winters compared to the other lakes. 

Report a problem or mistake on this page
Please select all that apply:

Thank you for your help!

You will not receive a reply. For enquiries, contact us.

Date modified: