The grounding line

Two glaciers are moving in two directions at once

Getz Ice Shelf's grounded ice is flowing 22.4% faster than in 2013. Over exactly the same years the shelf in front of it went 15.0% slower. Both slopes clear twice their own standard error, so neither is noise — they are two real signals pointing opposite ways, 37.4 percentage points apart. Averaged into one figure across the whole glacier they cancel, which is how a glacier whose grounded ice is accelerating gets filed under no detectable change.

This is the distinction most coverage of Antarctic ice loses. The grounded number is the one that bears on sea level; the shelf number mostly records what the shelf is doing after it has lost buttressing or shed a berg. They are worth reporting together and worth never adding up.

Every glacier where both domains can be measured

Indexed to 100 in the first year the two share. The solid line is grounded ice, the dashed line the floating shelf. Axes are per panel, not shared — the ranges differ by more than an order of magnitude — so the dashed rule at 100 is the anchor in each. The 6 glaciers where neither domain moves beyond its own uncertainty are in the table below rather than drawn.

Why the years have to line up

Each domain is measured over whatever years it has usable coverage for, and those windows do not always match. Thwaites Glacier is the case in point: its shelf series runs 2007–2022 and reads +50.5%, while its grounded series only starts in 2013. Refit over the 10 years the two actually share, that +50.5% becomes -0.1% ± 15.6 — inside its own error bars. The change is a step that happened before the grounded record begins, not a trend running alongside it. Subtracting the two published figures would have manufactured a gap of 42 percentage points out of a difference in start dates.

Every figure on this page is refit over the shared window for that reason. Where the windows already matched, the numbers are unchanged from the ones on the glacier pages.

All 11, measured over the years they share

How this is measured

Both domains refitted over only the years each glacier has usable coverage in BOTH, so the two percentages describe the same stretch of time and their difference means something. Each change is a least-squares slope over the span, expressed against the series mean, with the standard error of that slope carried alongside: a ±2σ bar that crosses zero is a change we cannot claim.

Coverage before 1999 is negligible and no continent-wide acceleration figure is defensible from this product. Per-glacier series are only reported where the coverage gate is met. Each box is a fixed window around the trunk, not a drainage basin, so these are speeds over a measured area and not mass-balance figures.

Grounded ice is landice AND NOT floatingice and the shelf is floatingice, using the masks carried by the velocity files themselves. Both masks are identical in every year, so neither can invent a trend. 2 of the 13 glaciers with a measured speed history have no floating trunk that clears the coverage gate, so they do not appear here at all; they are on the glacier directory with a grounded figure only.

See the whole ice sheet move

The same measured velocity field, with 262,000 particles advected through it — every glacier on this page is in there. The per-glacier histories are on the glacier directory, and today's ranked dispatches are on the wire.