The great ice sheets of the Northern Hemisphere at 18 moments of the last 3.6 million years, from the reconstructions of Batchelor and colleagues (2019): how big they grew, how far south they reached, which of today's cities and countries lay beneath them, and the land that appeared when the sea fell.
Twenty-six thousand years ago a sheet of ice lay over almost all of Canada and the northern United States, and another over Scandinavia, the Baltic and much of Britain. It was not the first time. The ice has advanced and retreated many times over the past few million years, and some of those advances reached further south than the last one.
In 2019 Christine Batchelor and eight colleagues gathered the published evidence for where the Northern Hemisphere's ice sheets reached, slice by slice through the Quaternary, and drew a best estimate of the ice for each of 18 time slices, from the late Gauss Chron, 2.6 to 3.59 million years ago, to the Last Glacial Maximum around 26,500 years ago. They also drew a minimum and a maximum outline for every slice before the LGM. I took the best estimates and asked what they cover today. Move the slider to step through the slices, and click anywhere on the map to see which of them reached that spot and how far it was from the ice edge.
Ice: the best-estimate outline for each time slice in Batchelor et al. (2019), Nature Communications 10, 3713 (doi:10.1038/s41467-019-11601-2), CC BY 4.0, generalised to about 7 km for the web. The outlines are drawn on today's coasts and relief, as the authors draw them, and give the extent of the ice, not its thickness. Sand: sea floor that a sea 60 m or 120 m lower than today would leave dry (ETOPO1 via ETOPO10, NOAA); the sea level is a fixed choice, not matched to each slice. Relief: AWS Terrain Tiles. Outlines and places: Natural Earth.
The ice we know best, the Last Glacial Maximum, covered 26.2 million km² in Batchelor and colleagues' best estimate, about 17.1 million of it on today's land and 9.0 million (34%) on what is now sea floor, from Hudson Bay to the Barents Sea. That is only the 4th largest of the 18 slices. The biggest was MIS 6, 132,000 to 190,000 years ago: 33.8 million km², 29% more than the LGM. MIS 12 (32.4 million) and MIS 16 (32.2 million) were close behind.
Ice area in each time slice
Area of the best-estimate outline, measured in an equal-area projection, newest at the top. Thin lines: the authors' minimum to maximum range for the slice; small ticks: the authors' own figure for the best estimate. Hatched: slices where the authors drew the ice at a warm peak, not at its largest. Hover a bar; click one to show it on the map.
The authors' own table of areas, which comes with the outlines, gives the same order for the seven largest slices, and its figures are within 4.2% of mine for every slice but one. The exception is 40 ka, where the outline holds 20% more ice than the table, mostly ice masses in north-east Asia. The minimum and maximum outlines show how uncertain the older slices are: for MIS 20 to 24 they run from 10.7 to 35.7 million km².
Three slices are smaller by design. For 45 ka, MIS 5a and MIS 5c the authors drew the ice at the warmest moment of the slice, not its largest, and these are the three smallest outlines, none above 7.2 million km². Add every slice together and the ice reached 37.0 million km² at one time or another, 41% more than the LGM alone. Only 4.8 million km², about half of it Greenland, was ice in every slice.
In North America the ice of the Last Glacial Maximum reached 39.0°N, near Bloomington, Indiana. In MIS 6 and MIS 12 it went further, to 37.2°N in southern Illinois, about 26 km from where the Ohio River flows into the Mississippi. In Europe the LGM ice reached its lowest latitude on the floor of the Celtic Sea south of Ireland, at 50.2°N; on the mainland it got no further south than 51.8°N. In MIS 6 a lobe pushed down the Dnieper valley in Ukraine to 48.3°N, near Dnipro, and in MIS 16 another reached 48.9°N in the Don basin of southern Russia.
The southernmost ice in each slice
Lowest latitude on any outline of 50,000 km² or more (the authors' size for an ice sheet), in North America (170°W to 50°W) and in Europe (12°W to 60°E). Hover a dot for the nearest city.
Mountain glaciers are not in these outlines. The authors drew only ice masses thought to have covered more than 50,000 km², so the ice caps of the Alps, the Pyrenees and the Rockies' southern ranges are left out, and the southern limits here are those of the continental ice sheets.
At the Last Glacial Maximum the ice covered 97% of Canada, all of Ireland, Sweden, Finland, Iceland, Estonia and Latvia, nearly all of Norway and Lithuania, 75% of Denmark, 61% of the United Kingdom and 21% of the United States. Across all the slices it reached 95% of Poland, 98% of Belarus, 60% of the Netherlands, 47% of Germany and 43% of Russia.
How much of each country lay under ice
Share of today's country area inside the LGM outline (dark) and inside at least one of the 18 outlines (pale). Countries with at least 5% in any slice and more than 20,000 km². Hover a bar.
Natural Earth lists 7,342 cities and towns around the world. 709 of them stand inside the LGM outline, 27 with a million people or more, and their population figures add up to about 141 million people, 5.9% of the total for all the places on the list. That is a rough guide, not a census, but it gives the scale. The biggest is New York, 13 km inside the edge, which matches the usual account that the ice reached Manhattan. Chicago lay 112 km inside, Toronto 190 km, Berlin 42 km and Seattle 17 km. Counting every slice, 1,050 places with about 230 million people were under ice at least once, among them Moscow, Philadelphia, Kyiv, Birmingham and Warsaw.
Big cities and the ice edge
Distance from the city centre to the edge of the LGM ice (negative: inside the ice), and the slices whose ice covered it, oldest on the left. Click a row to see the city on the map.
London was never under the ice in any of these outlines, but in MIS 12, the Anglian glaciation of British geologists, the edge came within 16 km of the centre. Moscow was 246 km beyond the LGM ice and under three earlier ones. Paris, Prague and Amsterdam stayed clear of the LGM ice, and of every other slice except, for Amsterdam, MIS 6.
The water in all that ice came from the sea. At the Last Glacial Maximum the oceans stood about 120 to 125 m lower than today, by the figures quoted on Wikipedia. On the ETOPO1 sea floor, a fall of 120 m uncovers 21.2 million km², 14.2% more land than the world has now. 3.7 million km² of that was itself under the ice, which leaves 17.5 million km² of new dry land. A fall of only 60 m already uncovers 67% as much.
The biggest new lands
Ice-free sea floor uncovered by a fall of 60 m and of 120 m, in boxes drawn round the shelf seas (my choice of boxes; see the methods). Hover a bar.
The largest new lands were in the tropics. The Sahul shelf joined New Guinea and Tasmania to Australia (2.9 million km² of new land in my box), and Sundaland joined Sumatra, Java and Borneo to mainland Asia (2.3 million). In the Arctic, the shelves of the Laptev, Kara and East Siberian seas added 1.8 million, and Beringia, the Bering and Chukchi seas, 1.6 million: along the meridian of the Bering Strait, dry land would have run unbroken for about 1,914 km from north to south. In Europe, the southern North Sea and the Channel became Doggerland, about 278,000 km² of ice-free land in my box; its northern part lay under the ice.
On a grid this coarse, about 18 km, every one of the 11 land bridges I tested is already made at 60 m: Britain to the continent, Ireland to Britain, Asia to North America, New Guinea and Tasmania to Australia, Sumatra, Java and Borneo to Asia, Sri Lanka to India, Taiwan and Sakhalin to the mainland. Narrow, deeper straits can hide below that resolution, so I leave out Japan and Newfoundland.
Ice. The best-estimate shapefiles from Batchelor et al. (2019), from the authors' repository on the Open Science Framework (osf.io/7jen3). The 18 time slices and their ages are the authors': the LGM (about 26,500 years ago), 30, 35, 40 and 45 ka, MIS 4 (58,000 to 72,000 years ago), MIS 5a to 5d (72,000 to 117,000), MIS 6 (132,000 to 190,000), MIS 8, 10, 12, 16, MIS 20 to 24 (790,000 to 928,000), the early Matuyama Chron (1.78 to 2.6 million years ago) and the late Gauss Chron (2.6 to 3.59 million). The files are in a polar Lambert equal-area projection, except the LGM, which is in latitude and longitude; I put everything in the equal-area projection, repaired invalid shapes, dissolved each slice to one outline and measured its area there. Two files, 45 ka and the late Gauss Chron, also carry today's glaciers worldwide, Antarctica included; I keep only the Northern Hemisphere parts.
What the outlines are. The authors stress that their reconstructions are generalised and partly subjective, that they should not be used for local studies, and that some, MIS 8 and MIS 10 above all, are highly uncertain and indicate the amount of ice rather than where its edge lay. For slices without direct evidence they borrowed the ice of a slice with a similar climate, and they used templates for Greenland, Iceland, the North American Cordillera and north-east Asia. Each outline aims at the largest extent within its slice, which may not have been reached everywhere at once. A city's distance to the edge is therefore a guide to the scale, not a local fact.
Places and distances. Countries are Natural Earth's (1:10m); their area under ice is measured in the same equal-area projection. Cities and towns are Natural Earth's populated places with their population figure (pop_max), which counts the built-up area; summing them is approximate. Distances to the ice edge are true distances on the WGS84 ellipsoid from the place to the nearest point of the outline. On the map, the distances are worked out in your browser from the generalised outlines and can differ by a few kilometres.
Sea floor. ETOPO10, a 10 arc-minute version of NOAA's ETOPO1, with cell areas on the sphere. Land today is ground above sea level or inside Natural Earth's land; sea is water joined to the oceans (the Black Sea is kept, the Caspian and lakes are not). Depth is the bedrock surface, so floating ice shelves do not count as shallow sea. The ice outlines are drawn on today's coasts, and I do the same: I do not model the crust sinking under the ice or rising since, or the sediment laid down, so the lower sea levels are a simple flood fill in reverse. The shelf boxes are mine.