feat(ice): world 39, two answers to one question a third apart - #59
Merged
Conversation
The NSIDC Sea Ice Index, both hemispheres, all twelve months since 1979, with every trend, error bar, rank, gap and percentage computed in the browser. The convention Extent counts a grid cell as ice if 15 percent of it is ice; area sums the fractions. So extent is always larger, at the September minimum by about a third, and almost every headline number ever published about sea ice is extent. The 15 percent is where the passive microwave signal becomes trustworthy, not a level at which the ocean changes character. The gap runs 35, 26, 32 and 35 percent across 1979, 1996, 2012 and 2025, so it cannot be subtracted away as a constant either. Also named: the hole in the data over the pole, which is assumed ice-covered and has shrunk as instruments changed. The window Arctic September ice declines 13.6 percent per decade ending at 2012, the record low year, 12.1 percent ending at 2024, and 11.9 percent carrying on to today. NSIDC publishes about 12.2, and ending where they ended reproduces it to a tenth. Ending on a record is what makes a slope look steepest, so every trend here carries its window and its standard error. September is falling several times faster than January, which a single annual figure hides. The two poles Antarctic February ice ROSE 0.13 million square km per decade to 2014 at 2.3 sigma, then fell 0.92 per decade after, at 2.5 sigma, seven times steeper in the opposite direction. The full-record trend containing both is 1.5 sigma, which is no signal at all: one series holding a real rise, a real fall, and a total indistinguishable from no change. The Arctic decline, for contrast, is 13 sigma. Quoting those two side by side as comparable claims is a statement about error bars nobody printed. An assumption corrected by measurement. The first version of the test file asserted the early Antarctic increase had never been significant and was only noise. Measured, it is 2.3 times its standard error and the winter rise is 4 times. The test now pins the measurement and the copy says what the data says. 36 tests validate against published values: the record low September of 2012 at 3.57 million square km, the record low Antarctic February of 2023, NSIDC's published decline reproduced on their window, area never exceeding extent across all 1,100-plus rows, and the 1987 to 1988 satellite outage left as nulls rather than zeros. Stated on the page: melting sea ice does not raise sea level because it already floats; this is area and not volume, and volume has fallen further; no ice-free date, because extrapolating a noisy 47-year series to a threshold needs physics; nothing before 1978, because the ship and aircraft records are not comparable and are not spliced on. Three defects a screenshot caught and assertions did not. The percentile band was drawn as a self-crossing hourglass, from reversing twice: once on the array of path commands and once on the percentile index, so x ran forward while y ran backward. A trend window was labelled "the Arctic's record low year" while showing Antarctic numbers. And its note asserted "steeper, because it ends on the record", which is true in the Arctic and false in the Antarctic, where the years since 2023 partly recovered; that note is now computed from the two slopes rather than assumed. The cron's "nothing changed" branch is real, not decorative: the fetch script compares every field except its own build timestamp before writing, so the repository does not collect a commit a month when NSIDC has published nothing. Verified: 1,459 tests pass, tsc clean, fresh production build, all 39 routes sweep clean, no console errors, no mobile overflow. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
1 Skipped Deployment
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The NSIDC Sea Ice Index, both hemispheres, all twelve months since 1979, with every trend, error bar, rank, gap and percentage computed in the browser.
The convention
Extent counts a grid cell as ice if 15 percent of it is ice; area sums the fractions. So extent is always larger, at the September minimum by about a third, and almost every headline number ever published about sea ice is extent. The 15 percent is where the passive microwave signal becomes trustworthy, not a level at which the ocean changes character. The gap runs 35, 26, 32 and 35 percent across 1979, 1996, 2012 and 2025, so it cannot be subtracted away as a constant either. Also named: the hole in the data over the pole, which is assumed ice-covered and has shrunk as instruments changed.
The window
Arctic September ice declines 13.6 percent per decade ending at 2012, the record low year, 12.1 percent ending at 2024, and 11.9 percent carrying on to today. NSIDC publishes about 12.2, and ending where they ended reproduces it to a tenth. Ending on a record is what makes a slope look steepest, so every trend here carries its window and its standard error. September is falling several times faster than January, which a single annual figure hides.
The two poles
Antarctic February ice ROSE 0.13 million square km per decade to 2014 at 2.3 sigma, then fell 0.92 per decade after, at 2.5 sigma, seven times steeper in the opposite direction. The full-record trend containing both is 1.5 sigma, which is no signal at all: one series holding a real rise, a real fall, and a total indistinguishable from no change. The Arctic decline, for contrast, is 13 sigma. Quoting those two side by side as comparable claims is a statement about error bars nobody printed.
An assumption corrected by measurement. The first version of the test file asserted the early Antarctic increase had never been significant and was only noise. Measured, it is 2.3 times its standard error and the winter rise is 4 times. The test now pins the measurement and the copy says what the data says.
36 tests validate against published values: the record low September of 2012 at 3.57 million square km, the record low Antarctic February of 2023, NSIDC's published decline reproduced on their window, area never exceeding extent across all 1,100-plus rows, and the 1987 to 1988 satellite outage left as nulls rather than zeros.
Stated on the page: melting sea ice does not raise sea level because it already floats; this is area and not volume, and volume has fallen further; no ice-free date, because extrapolating a noisy 47-year series to a threshold needs physics; nothing before 1978, because the ship and aircraft records are not comparable and are not spliced on.
Three defects a screenshot caught and assertions did not. The percentile band was drawn as a self-crossing hourglass, from reversing twice: once on the array of path commands and once on the percentile index, so x ran forward while y ran backward. A trend window was labelled "the Arctic's record low year" while showing Antarctic numbers. And its note asserted "steeper, because it ends on the record", which is true in the Arctic and false in the Antarctic, where the years since 2023 partly recovered; that note is now computed from the two slopes rather than assumed.
The cron's "nothing changed" branch is real, not decorative: the fetch script compares every field except its own build timestamp before writing, so the repository does not collect a commit a month when NSIDC has published nothing.
Verified: 1,459 tests pass, tsc clean, fresh production build, all 39 routes sweep clean, no console errors, no mobile overflow.
What & why
How to test
Checklist
npm run buildpasses locally.npx vitest runpasses (added or updated tests for new logic where it makes sense).docs/DATA_SOURCES.mdin this same PR..envconfig (a new source needing a key ships with a keyless fallback).lib/geo.ts; the globe mesh is not rotated (coordinate convention is locked).useFrameloops; heavy scenes stay dynamically imported (60fps budget on integrated graphics).The honesty rule
This project's credibility rests on one rule: every number on screen traces to a real data source or a documented calculation, and anything simulated is labeled as simulated. By opening this PR you confirm your change respects it.
Screenshots / recording