Lake Ice In Dates: When Lakes Typically Freeze Over

Q: How do I know whether the ice is safe to walk on?

Contact your state natural resources agency and look at the water yourself. Every figure here is the day a lake closed over in a past winter, and the days right after a lake closes are the days it most looks frozen and least is.

The short answer

November 15 to February 10

Across the 40 water bodies with a freeze record, the typical closing runs from November 15 on Maple Lake in WI to February 10 on Grand Traverse Bay in MI. That is 87 days inside five states, and most of it is depth and surface area rather than latitude. Every figure on this page is a median drawn from past winters, not an observation of any lake now.

87 days

Between the earliest and latest typical freeze

17

Records spanning a century or more

109

Winters with no freeze at all, on 11 water bodies

5

States with a freeze record, against nine for the thaw

Every lake with a freeze record, in the order they close

Read this from the top and you are reading the country freezing over. The small shallow lakes of northern Wisconsin and Minnesota go first, in the middle of November. The deep ones follow through December. The Great Lakes bays come last and sometimes do not come at all.

Each bar runs from that water body's earliest recorded closing to its latest. The solid dot is the median across the whole record and the ring is the median across its most recent 30 recorded winters. The scale starts near the earliest date in the set rather than at January 1, because these are dates a few weeks apart and a full year axis would hide the whole picture.

Overhead view of pancake ice on Lake Superior, rounded plates of new ice with raised white rims and pale turquoise centres packed edge to edge across the water — Pancake ice on Lake Superior. Each plate froze on its own, then collided with its neighbours until the collisions rounded and rimmed it. A closing date is the day all of this finally welds together, which on a small lake takes a night and on water this size can take a month or never happen at all. “Pancake Ice on Lake Superior” — National Park Service, Pictured Rocks National Lakeshore — Public domain

Ice in and ice out are timing questions, answerable from records that already exist: every figure on these pages is a date somebody observed in a past winter. Ice strength is a different question, and the state agency linked on each state page is the source for the current season.

What two centuries show

The ice season is closing from both ends

This is the part the record shows directly, and it is the reason the freeze half of the record is worth publishing separately. 22 lakes here carry a least squares slope on both variables. On 21 of them the freeze has been arriving later, and on 21 the thaw has been arriving earlier. On 21 it is both at once.

The median freeze shift across those lakes is 7.1 days later per century. The median thaw shift is -5.6 days per century. Add one lake’s own freeze line to its own thaw line and the median lake in that set has lost about 12.3 days of ice season per century. The widest is Lake Superior at Bayfield, where the freeze has moved 15.4 days later and the thaw 16.2 days earlier, which is 31.6 days of ice season per century on a record that has run since 1858.

Every one of those figures is a straight line fitted to one lake’s own observations, which is the standard description in the ice phenology literature. It is arithmetic on dates that were written down. It is not a projection, there is no model underneath it, and it does not say what the next decade will do.

Freeze arriving later
Thaw arriving earlier

Each bar is one lake's own freeze line added to its own thaw line, both fitted by least squares to that lake's observations and both already published in the source data. The figure on the right is how many days of ice season that adds up to per century. It is arithmetic on a record, not a model and not a projection, and only lakes with 80 or more years on both variables appear.

A wall of shelf ice rising above a lake surface strewn with broken plates of new ice and patches of dark open water, under a blue sky with high cloud — Shelf ice standing above a field of broken new plates, with open water still showing between them. This is the state a lake is in for days or weeks before anybody writes a closing date against it, and it is why two observers on the same shore can pick two different days. “Ice along Lake Superior” — National Park Service, Pictured Rocks National Lakeshore — Public domain

Where the records are

The five states with a freeze record

Of the 116 United States water bodies in the international database with 40 or more usable observations, 40 carry a freeze series, and they sit in five states; nine states have a thaw record. The thaw was the date that paid: it set when the boats went in, when the sporting camps opened, when the ferries ran. The freeze mattered to fewer people, so fewer people wrote it down.

State	Water bodies	Record starts	Deepest record	Typical ice in there	Earliest to latest in the state
Wisconsin	17	1851	Lake Monona (166 winters)	Dec 15	Nov 15 to Jan 9
New York	12	1816	Otsego (167 winters)	Jan 11	Dec 2 to Feb 2
Michigan	4	1852	Grand Traverse Bay (122 winters)	Feb 10	Dec 6 to Feb 10
Minnesota	4	1860	Detroit (106 winters)	Nov 22	Nov 22 to Dec 2
Iowa	3	1916	East Okoboji Lake (92 winters)	Nov 30	Nov 30 to Dec 12

The long ones

The deepest freeze records in the country

These two have between them more than three hundred recorded winters, and neither record was started for science. Otsego was kept by a lakeside observer and a university field station; the Madison lakes were kept by a chain of people beginning at the Washburn Observatory who simply wrote the date down every year.

Otsego

Lake in New York · 167 recorded winters, 1850 to 20193 winters open

Jan 11

typical ice in

Jan 19

last 30 recorded winters

Dec 13

earliest, 1917

Feb 25

latest, 1932

Across the whole record the closing date has moved 7 days later per century across the record.

Its ice out has moved 5 days earlier per century across the record, so adding the two lines together this lake’s ice season is about 12 days shorter per century.

3 winters on record it never closed over at all, most recently the one ending in 2017. Those winters are counted separately from the 167 dates above.

Mean closing date per decade. Only decades with four or more observations appear, and the scale starts near the shortest bar rather than at January 1 so that small differences stay visible rather than collapsing into one length.

Lake Monona

Lake in Wisconsin · 166 recorded winters, 1851 to 2018

Dec 15

typical ice in

Dec 24

last 30 recorded winters

Nov 23

earliest, 1857

Jan 30

latest, 1932

Across the whole record the closing date has moved 8 days later per century across the record.

Its ice out has moved 12 days earlier per century across the record, so adding the two lines together this lake’s ice season is about 20 days shorter per century.

Mean closing date per decade. Only decades with four or more observations appear, and the scale starts near the shortest bar rather than at January 1 so that small differences stay visible rather than collapsing into one length.

Grand Traverse Bay did not close over at all in 44 winters on record, which is 27 percent of the winters in its span, and the eight most recent of those ended in 2006, 2007, 2008, 2010, 2011, 2012, 2016, 2017. The source records those winters rather than leaving them blank, and this page keeps them, because dropping them and averaging only the winters the water did close would bias the record toward an earlier freeze. In Michigan that is the single most important thing the freeze record has to say.

Satellite view of Lake Superior in mid December with the surrounding land under snow, the open lake dark, and pale new ice showing in the shallow western bays — Lake Superior on December 15, with the first ice of that winter forming in the shallow western bays while the deep middle stays open and dark. The Bayfield record on the Wisconsin shore is drawn from exactly this corner, and its typical closing date is January 1. Grand Traverse Bay, the latest typical freeze in this dataset at February 10, is also the water body most likely to record nothing at all: in 44 of its winters it never closed across. “Lake Superior (MODIS 2018-12-20)” — MODIS Land Rapid Response Team, NASA GSFC — Public domain

Other states

States without a freeze record

5 states have a freeze record and a page here; the ice out side covers eight. The difference is the first entry below: Maine, New Hampshire and Massachusetts have long lake ice records on this site that contain no freeze dates at all. The rest are listed with the reason.

Maine, New Hampshire and Massachusetts

All three have a lake ice-out page on this site, and their records are thaw records. The USGS compilation behind the New England thaw dates is an ice-out report: it collected the day the ice went off 29 lakes and not the day it came on, and no New England water body in the international database carries a freeze series of 40 or more winters. The thaw was the date that mattered commercially, because it set when the boats went in, so the thaw is the date somebody wrote down.

Vermont

Vermont shares Lake Champlain with New York and shares the longest freeze record in the country with it. The international database files that series under New York, so that is where it appears here, and the reader who wants it should follow the New York page. No lake wholly inside Vermont carries a freeze series in either compilation this site is built on.

Alaska

Alaska has one long freeze-up and breakup record in the database and it is the Tanana River at Nenana, watched since 1917 by the Nenana Ice Classic. A river freezing up is a different physical event from a lake closing over, it is driven by current rather than by heat loss from a still surface, and it does not belong in a lake ice-in table. It is a magnificent record and it deserves its own treatment.

The Dakotas, Montana, Idaho, Wyoming, Colorado, Washington

Cold enough for lake ice every winter, and no lake in either compilation. Records exist locally in places, held by lake associations, resorts and individual observers, and none of them has been gathered into a published series.

Everywhere else

Roughly two thirds of the country either does not reliably grow lake ice or has never had anyone write down the date it arrived, so no freeze record exists for it, and the states above are the ones where somebody kept one.

Ice in is the autumn halfThe day a lake closed over in a past winter. Five states, 40 water bodies, 3,402 recorded freezes. You are on it.

Ice out is the spring halfThe day the ice finished leaving in a past spring. Nine states, 116 water bodies, and the oldest single observation in the country, from 1807.

Provenance

Where all of this came from

40 water bodies, 3,402 recorded freezes

Every figure on these pages is derived from the international ice phenology database held at the National Snow and Ice Data Center. These pages cover every United States water body in it with 40 or more usable observations, and the medians, the extremes, the decade means and the least squares trends are computed from those observations. The same generated file feeds the ice out side of the site.

Benson, B., Magnuson, J. and Sharma, S. (2000). Global Lake and River Ice Phenology Database. (G01377, Version 1). Boulder, Colorado USA. National Snow and Ice Data Center. https://doi.org/10.7265/N5W66HP8 Accessed 2026-08-24.

Definitions of ice in genuinely differ between lakes and between observers, and the source says so about its own data. Minnesota requires the whole lake frozen with the cover holding; Wisconsin uses the first date the majority of the surface is estimated to be frozen; the Bayfield series on Lake Superior uses the last boat out of the harbour. Several New York series were re dated by a later compiler to mean first freeze rather than final closing. That matters most when comparing two lakes against each other, and least when comparing one lake against its own past.

Common questions

Lake ice in FAQ

What is a lake ice in date?

It is the day in autumn or winter when a lake closed over with ice. The test differs from lake to lake and always has. Minnesota DNR uses the whole lake being frozen over with the cover holding through the winter. The Wisconsin State Climatology Office uses the first date the majority of the surface is estimated to be frozen. At Bayfield on Lake Superior the date is the day the last boat left the harbour, because nobody can watch a lake that size close. Those are genuinely different thresholds, which is why comparing a lake against its own past is far more useful than comparing two lakes against each other.

How do I know whether the ice is safe to walk on?

When do lakes usually freeze in the United States?

There is no single answer and the range in this data is the reason. Across 40 water bodies with 40 or more recorded winters, typical freeze dates run from November 15 on Maple Lake in Wisconsin to February 10 on Grand Traverse Bay in Michigan. That is 87 days of spread inside five states, and most of it is depth and surface area rather than latitude. A small shallow lake gives up its summer heat in weeks. A deep one takes months, and a bay of a Great Lake may not manage it at all.

Why do only five states have pages?

Because five states are all that have a freeze record. The international ice phenology database holds 116 United States water bodies with 40 or more usable observations, and only 40 of them carry a freeze series at all. Those 40 sit in Wisconsin, New York, Michigan, Minnesota and Iowa. The ice-out side of this site covers nine states: the freeze is the half of the ice season fewer people wrote down.

Is ice in happening later than it used to?

In this data, overwhelmingly yes. Of the 22 lakes here that carry a least-squares slope on both variables, 21 show the freeze arriving later and 21 show the thaw arriving earlier. The median freeze shift is 7.1 days later per century and the median thaw shift is -5.6 days per century. Adding one lake's own freeze line to its own thaw line, the median lake in that set has lost about 12.3 days of ice season per century. These are lines fitted to observations, which is the standard description in the ice phenology literature. They describe what each record did. They are not projections and there is no model behind them.

What does it mean when a lake does not freeze at all?

It means exactly that, and a good record writes it down rather than leaving a blank. Grand Traverse Bay has 44 winters on record in which it never closed over, the most of any water body here, and across this whole set there are 109 such winters on 11 lakes. Those winters matter more than the dates do. If you drop them and average only the winters a lake did close, you quietly bias the record toward an earlier freeze, because the winters warm enough to prevent a freeze are exactly the ones that would otherwise have produced the latest dates.

Where do I find whether a lake has frozen this winter?

From the state agency. Minnesota DNR runs a public ice-in map every autumn and is the only state that does. The Wisconsin State Climatology Office publishes the Madison lakes each winter. The National Weather Service in Burlington keeps the Lake Champlain closing record and adds to it annually. NOAA maps ice cover on the Great Lakes themselves. Each state page links its own source; these pages hold the long record.

Why is there no forecast of this winter's ice in date?

Most of these records end between 2007 and 2020, so they describe past winters rather than predicting the next one. Each state page shows how the current point in the season has played out across the recorded winters, and the state agency linked on every page publishes the current winter as reports arrive. Nothing on these pages describes ice condition or ice strength; for anything involving going out on ice, the state agency is the source.

How is ice in different from ice out?

They are the two ends of the same season and they are not mirror images. Ice out is a thaw, driven by sun angle and rain as much as by air temperature, and it happens over days on a sheet that is already rotten. Ice in is a heat loss problem: a lake has to shed the warmth it stored all summer before its surface can hold, which is why depth predicts the freeze date far better than it predicts the thaw. In this dataset the freeze has moved further than the thaw has, and it is also the half that fewer observers recorded, so there is less of it to look at.

Satellite view of snow covered northern New York in mid January, with the darker Adirondack uplands in the centre and the long narrow shape of Lake Champlain running dark and open down the right hand side — Northern New York on January 20, snow across the whole scene and Lake Champlain still a dark open ribbon down the right. The National Weather Service has kept Champlain’s closing record in Burlington since 1816, and in 42 of those winters the answer it recorded was that the lake never closed at all. “Winter in the Adirondacks (MODIS 2022-01-22)” — MODIS Land Rapid Response Team, NASA GSFC — Public domain

All dates on these pages are historical observations, not forecasts. The most recent observation anywhere in this data is from 2020; for the current winter, each state page links to the agency that publishes it. These pages carry no live measurement and no statement about ice condition or ice strength, and no figure here says a lake is frozen; for anything involving going out on ice, the state agency is the source. Median, earliest, latest, decade mean and change per century are computed from the observations. The other end of the season: lake ice out dates. Related: winter outlook, first snow, lake guides.
Hero image: “Ice on Green Lake” by Wideshpan6, CC BY-SA 4.0
Deepest freeze record here, Otsego: typical ice in January 11 across 167 recorded winters.