DSC 106 · UC San Diego · Final Project

WATCHING THE ICE MELT

A satellite record of polar ice moving from seasonal rhythm to structural loss: Arctic sea ice shrinks, Antarctic summer ice breaks pattern, and the ice sheets lose mass year after year.

Joey Sandoval · Scarlett Scott · Roberto Huizar
NSIDC SEA ICE INDEX V4 · NASA EARTHDATA MODIS · GRACE / GRACE-FO · 1978 – 2026
Scroll to begin ↓
01 · The number

September 16, 2012.

The Arctic Ocean's sea ice cover hits a number nobody had predicted.

3.34 million km²

The lowest September minimum ever recorded. Roughly half of what was considered normal a generation earlier.

This is not just a single bad year, but a signal that the old Arctic had become hard to find.

02 · The baseline

What "normal" used to look like.

In the 1980s, the average September minimum sat near 7.0 million km². The ice grew and shrank with the seasons, but the floor stayed steady year after year.

The old floor matters because sea ice is supposed to be seasonal with normal winter growth and summer melt. However, a lower summer floor means the Arctic is entering autumn with less reflective ice left to rebuild from.

03 · The drift

Then the line started slipping.

Through the 1990s and 2000s, scientists watched the September minimum erode more and more each year. Every individual season looked noisy, and the average kept dropping.

Not only does the minimum continue dropping over the years, but the whole cloud of years shifts downward as a whole, turning former extremes into ordinary seasons.

04 · The wake-up

September 2007.

The Arctic ice area dropped to 4.15 million km², a million below the previous record. Models previously had said that this would not happen until mid-century.

The climate community started using the phrase regime change. The record was not just low; it revealed how much thinner, younger, and more fragile the remaining ice had become.

05 · The record

Five years later, the floor fell out again.

2012's September minimum bottomed at 3.34 million km². Open water reached almost to the pole for the first time in human history.

Since then, no year has recovered to even the late-1990s "normal." The Arctic did not bounce back after a shock, instead it settled into a lower range.

06 · Step back

Decade by decade, the story is the same.

Smoothing out the noise, the average September minimum has fallen from 6.96 → 4.38 million km² across five decades.

The 2010s and 2020s are essentially flat, at a new baseline that would have looked extreme in the early satellite records.

07 · The other pole

Then, Antarctica followed suit.

For forty years the Antarctic had been stable or even gaining sea ice. Then in February 2023, the summer minimum collapsed to 1.85 million km², far below any previous year on record.

The Arctic tells the long story of decline. Antarctica shows how quickly a stable-looking system can slip into new territory.

That contrast shows how climate change does not always arrive as the same shape everywhere.

08 · The land ice

The ice sheets are losing mass too.

Sea ice changes the planet's reflectivity. Land ice changes sea level. GRACE and GRACE-FO gravity measurements show Greenland and Antarctica together reaching about -7.65 thousand gigatonnes below the early satellite baseline by 2020.

That means the sea-ice story is not isolated. The cryosphere is losing both floating cover and land-stored ice.

09 · The heat

Warming is strongest when the Arctic should be refreezing.

In this heatmap, each row is a month and each column is a decade. The darker red cells show larger temperature anomalies. Notice that the strongest warming appears in winter and late fall, not only during summer melt season.

That pattern matters because sea ice normally regrows in the cold months. When fall and winter are warmer, freeze-up starts later, the ocean stays exposed longer, and the next melt season begins from a weaker starting point.

10 · The feedback

Less ice changes how much sunlight the Arctic absorbs.

This diagram explains the mechanism behind the line charts. Bright sea ice reflects much of the sun's energy. Dark open ocean absorbs more of it. When ice melts, the surface gets darker; when the surface gets darker, it absorbs more heat; and that extra heat can melt more ice.

The important observation is that this is a reinforcing loop. The loss of ice does not just show warming; it can also help amplify warming.

11 · The horizon

Future emissions change the timing of an ice-free Arctic.

Each line represents a different emissions pathway. The red line is a high-emissions future where warming pressure stays strong, so the Arctic crosses the ice-free threshold earliest. The orange line is a middle pathway where sea ice still keeps declining, but the crossing happens later. The blue line represents stronger emissions cuts, where the decline slows and the threshold is delayed the longest.

This projection chart asks what happens next: do emissions keep pushing that baseline downward quickly, or do cuts slow the movement enough to delay the first ice-free September?

12 · Your turn

Now explore the data yourself.

The full interactive section below lets you check each part of the story yourself: decade averages, individual years, ice-sheet mass loss, temperature anomalies, future scenarios, and the albedo feedback loop.

As you explore, look for the same question in every view: is the chart showing one unusual year, or is it showing that the baseline has moved?

Section II

Explore the data.

Six interactive charts using downloaded NSIDC v4 sea-ice CSVs, GRACE/IMBIE ice-mass data, and NASA Earthdata MODIS products. Drag the slider, click a decade, hover any bar.

01 · Arctic September Minimum Sea Ice Extent (Decadal Small Multiples)
Real NSIDC v4 decadal averages. Click any decade to see how far below 1980s baseline it sits.
Click a decade.
NASA MODIS MOD29 Sea Ice product (daily 1 km resolution) provides ground-truth validation for NSIDC passive-microwave extent estimates.
02 · Arctic Sea Ice: Daily Extent, All Years 1978–2026
Drag the slider to spotlight any year against the 49-year envelope.
YEAR2012
MODIS Terra Band 1–2 false-color composites visually confirm ice extent boundaries flagged by NSIDC passive-microwave algorithm.
03 · Ice Mass Loss: Greenland + Antarctica (2002–2020)
Hover any bar for cumulative gigatonnes lost in the GRACE / GRACE-FO composite.
GreenlandAntarctica
Hover a bar.
GRACE / GRACE-FO mass-balance values come from the IMBIE / NASA JPL Tellus composite CSV; downloaded MOD10A1 snow-cover granules provide surface-ablation context.
04 · Arctic Temperature Anomaly: Month × Decade
Hover any cell to read month, decade, and anomaly °C above the 1981–2010 mean.
Hover the grid.
NASA Earthdata MOD11A2 land-surface-temperature granules were downloaded for Arctic land context; the heatmap summarizes the warming pattern used in the story.
05 · When Is the First Ice-Free Arctic Summer?
Click a scenario button to toggle its model envelope on or off.
Projection curves are scenario guides anchored to the observed NSIDC v4 decline; downloaded MOD29 granules provide sea-ice product context for the 15% ice threshold.
06 · The Albedo Feedback Loop
Click any node to read what MODIS is measuring at that step in the loop.
Click a node.
Downloaded MCD43A3 BRDF/Albedo granules connect the sea-ice extent story to measured reflectance change at 500 m.

Project Writeup

1. What data is driving the final story?

The project now uses downloaded NSIDC Sea Ice Index v4 CSVs for daily Arctic and Antarctic sea-ice extent, a public GRACE / GRACE-FO ice-sheet mass-balance CSV for Greenland and Antarctica, and NASA Earthdata MODIS granules for sea ice, snow cover, land-surface temperature, and albedo context. The central story is that the Arctic did not lose ice in one dramatic moment; it lost its old baseline. The 1980s September minimum averaged about 6.96 million km², the 2010s averaged about 4.43 million km², and the 2020s remain near that lower floor. The 2007 crash made the shift visible, 2012 set the record low at 3.34 million km², and Antarctica's 2023 summer minimum shows that the other pole can also break from its historical pattern.

The sea-ice record is useful because it is long enough to separate weather from climate. A cold summer or cloudy melt season can still produce a higher year, and a storm can help break up ice in a lower year. But when the full satellite record is shown together, the noisy seasonal lines form a clear envelope: the lower edge of Arctic summer ice has moved downward. That is why the story begins with one dramatic number, then immediately asks the reader to compare that number against the old baseline.

The Antarctic chart adds a second kind of evidence. Unlike the Arctic, Antarctic sea ice did not show a simple long-term decline for most of the satellite era. That makes the 2023 summer minimum more striking. It is not presented as proof that both poles behave identically; it is evidence that polar systems can hold a pattern for decades and then shift quickly. The contrast between the Arctic's gradual loss and Antarctica's sudden break gives the piece its larger structure.

2. What should readers take away?

The charts are designed to move from evidence to mechanism. The scroll story first establishes the sea-ice record, then the Explore section lets readers test the pattern by decade, by year, by ice-sheet mass, and by feedback loop. GRACE shows that the sea-ice story is not isolated: by 2020 the combined Greenland and Antarctic ice-sheet anomaly reaches roughly -7.65 thousand gigatonnes. MODIS connects the visual mechanism behind that loss: less bright ice and snow means less reflected sunlight and more absorbed heat. The point is not that every year is a record; the point is that the baseline itself has moved.

The project also separates three related but different cryosphere signals. Sea ice is floating ocean ice, so its retreat changes reflectivity and ocean heat absorption. Ice-sheet mass is land ice, so its loss contributes to sea-level rise. MODIS albedo and temperature products describe the surface conditions that help explain why ice loss can reinforce itself. Showing these together helps the reader avoid treating "melting ice" as one vague idea. The Arctic sea-ice line, the GRACE mass bars, and the albedo loop are different measurements pointing toward the same climate pressure.

The intended ending is not despair; it is agency with evidence. The projection chart is deliberately placed after the historical record because the future should not feel abstract. By the time readers reach the scenario buttons, they have already seen what a moving baseline looks like. The question becomes less "is the Arctic changing?" and more "how far do we let the new baseline move?"