Snow Hill Island
Snow Hill Island is a known penguin breeding site in the Northeast Antarctic Peninsula, at 64°31'S 57°27'W. Emperor penguins have been recorded here. Emperors breed on sea ice rather than land, which makes this one of the sites whose very surface disappears in a bad ice year. The sea around it has lost ice at 1.09 points a decade since 1979, which is inside its own uncertainty.
Species recorded here
The sea ice around this colony, 1979–2024
Averaged over 36 ocean cells within 100 km, the April–November sea-ice concentration here has moved -1.09 ± 1.25 points per decade — fitted 87% in 1979 to 82% in 2024. That is no detectable change. The slope does not clear twice its own standard error, so the honest answer here is that the ice around this colony has not measurably moved in 46 years. Over the fledging window it reads -2.39 ± 5.40.
Two limits, and both matter. First, this is not the fast ice the birds breed on. At 25 km the land-attached ice emperors breed on is narrower than one cell and cannot be resolved. This is the sea around the colony, not the platform under it. Second, the archive ends 2024-12-31 — it is a climate record, not a current reading, and the live sea-ice figure on the wire is a different product entirely.
Is this an independent measurement?
90 other known breeding sites lie within 200 km of this one, and the cells this measurement averages over reach 100 km — so this record and theirs are not independent readings of the sea. Thinned so that no two overlap, the 728 measurable sites in this archive become 56 places.
In the 2023 ice year — the last one this archive holds — the sea ice here lasted 291 days, which is this site's 3rd shortest season of 41 and below its own median of 361. It was below the median at 657 of the 728 sites measured here and at 44 of the 56 independent places, which is the most widespread short season in the record: the whole archive is drawn in one frame.
And when that ice goes, 1979–2023
Across 28 usable ice years the retreat date here has moved -9.8 ± 11.9 days per decade — fitted 2 Jan at the start of the record to 21 Nov at the end, a shift earlier of 42 days. That is no detectable change. The slope does not clear twice its own standard error, so the honest answer here is that the date has not measurably moved. Since 2000 the sea here has opened before 15 December — the point in the breeding calendar at which a chick still cannot swim — in 10 of 16 ice years. On the stricter 15% threshold — the day the sea around it opens outright rather than falling below its own normal — 18 of the 28 ice years never clear at all, so the figure above is the one to quote.
Two uncertainties, not one. The fit above carries ±11.9 days per decade of its own. The archive carries a second one that no other site prints: before 1988 the satellite flew every other day and the producer filled the missing ones, so 1,563 of 3,287 daily readings at this colony in that era — 48% — were interpolated rather than observed, and nothing in the data says which. Propagated through this particular fit, that is worth ±2.2 days per decade — a fraction of the fit's own. Both are printed; neither is folded into the other.
This is the day the sea around the colony opened, not the day the fast ice under the birds broke out — a 25 km cell cannot see one bay. An early exit here is a plausible companion to a bad season and never a cause of one. See all 61 colonies dated this way →
And how long that ice lasts, 1979–2023
Across 28 usable ice years the sea-ice season here averages 263 days and has moved -14.0 ± 14.5 days per decade — fitted 296 days at the start of the record to 235 at the end. That is no defensible change. On the stricter of the two gates this site uses the season length here has not measurably moved. In 9 of 41 ice years the sea here was still above its own normal when the ice year opened, so the freeze-up date is censored at the start of the year rather than measured.
The two ends add to that figure exactly, because within one ice year the season length is the retreat date minus the freeze-up date: -9.8 days per decade from the ice going, -4.3 from the ice arriving. See all 61 colonies measured this way →
And a second uncertainty, of a different kind. Before 1988 the satellite flew every other day and the producer filled the missing ones: 1,563 of 3,287 daily readings at this colony in that era — 48% — were interpolated rather than observed, and nothing in the pixels says which. Measured against a ground truth and propagated through this particular fit, that is worth ±2.7 days per decade against the fit's own ±14.5. The two ends carry their own: ±2.2 days per decade on the ice going and ±1.2 on the ice arriving. Those three figures are not added together — the trend decomposes exactly, but the uncertainties behind it were each measured over a different set of ice years, so each belongs to its own slope.
Why there is no count on this page
Penguins are too small to resolve from orbit. What is visible is the guano stain a colony leaves on the ice, which is why counting from satellite is a colour problem rather than a counting problem — and it is not solved. Our own detector correlates at r = −0.16 against ground counts across eight colonies, which is no relationship at all. Until that is fixed, this site publishes locations and says nothing about numbers. A number we cannot defend would be worse than none.