CMIP6 climate projections · DSC 106 final project
Climate change is usually communicated as a global average. So we broke the same data the IPCC uses down to the city level. Pick a place. Watch what climate change does, not to the planet, but to your weekend in 2080.
why this matters
first, a word on what we're showing
Climate scientists don't predict the future — they project what would happen if humanity follows different paths. The IPCC calls these Shared Socioeconomic Pathways, or SSPs. Each combines assumptions about energy, population, policy and growth.
The observed record from 1850 to 2014. Real measurements, not projections. The Earth warmed about +1.0 °C in this window.
We get to net-zero by mid-century, shift to clean energy fast, lower inequality. Warming stabilizes near +1.8 °C. This is the "if we act" path.
Current policies, modest progress, the world we're trending toward. Warming reaches about +2.7 °C by 2100.
No serious decarbonization. Economic growth powered by coal and oil through 2100. Warming hits +4.4 °C and keeps rising past 2100.
The "2.6", "4.5", and "8.5" in the names are the radiative forcing in watts per square meter at 2100 — basically, how much extra energy the atmosphere is trapping. Bigger numbers = more warming.
the global picture
Global mean surface temperature, 1850 to 2100. The four lines are the futures we get under four emissions pathways — from acting decisively to not acting at all.
The historical record. Real measurements. The Earth warmed about +1.0 °C — and the warming accelerated after 1980.
If we get net-zero by mid-century, warming flattens out near +1.8 °C. We just barely overshoot the Paris target.
Current policies, modest action. Warming reaches about +2.7 °C by 2100 — the world we're trending toward.
No serious decarbonization. Warming hits +4.4 °C — and keeps rising past 2100. The point of "this is what we avoid".
+1.5 °C is the Paris Agreement target. We've already crossed +1.0. Whether we hold the line depends on which path above we end up on.
make it personal
36 cities across every inhabited continent. The five visualizations below all respond.
the main story
Right now, San Diego has roughly — days a year that climb above 35 °C / 95 °F — what the heat index calls "dangerous."
Under SSP5-8.5, that climbs to — days. A —× increase. Effectively all of summer.
Every square is a day. Red squares are dangerously hot ones. Today, they're scattered across the summer. By 2080, they bleed into spring and fall — and the heart of summer becomes a solid wall.
Today, the worst heatwave above 35 °C runs about — consecutive days. By 2080 (worst case): —. In Delhi: 270 days in a row.
Days above 30 °C are common today; days above 40 °C (104 °F, heatstroke territory) are rare. By 2080, both multiply — and the 40° line, near-zero today, becomes a recurring event.
Lagos goes from essentially never to 110 days a year. Madrid: 0 → 49. Delhi: 111 → 179. The cities at the bottom of the list still triple their hot-day counts.
the underlying warming
The historical record at left, then the projections fork after 2015. 8-model ensemble; shaded bands show p10–p90 model spread.
Under SSP1-2.6, your city's temperature flattens out within decades — warmer than today, but stable. This is the version of the future we have to fight for.
SSP2-4.5 keeps climbing. Mid-century looks recognizable; end-of-century starts to feel like a different climate zone.
Under SSP5-8.5, the line keeps climbing past 2100. The gap between this red line and the blue one above is what your generation decides.
the same city, three futures
Mean annual temperature change in San Diego, from 1950–2014 baseline to 2071–2100, under each of three emissions pathways.
climate analog
In 2080, San Diego will feel like
—
polar amplification
Projected mean annual temperature change, end of century under SSP5-8.5. The cities that warm most aren't on the equator — they're high-latitude and continental: Montreal, Moscow, Toronto. Polar amplification in action.
extreme heat days, ranked
Days per year above 35 °C today vs end of century. Lagos goes from essentially never to more than 100 days a year. Click any row to switch the explorable to that city.
the whole world, mapped
Projected change in annual mean temperature, 2071–2100 vs 1950–2014 baseline. Toggle between temperature and precipitation; drag the slider to morph between emissions pathways.
We pulled the CMIP6 climate-model archive directly from the
Pangeo cloud mirror (gs://cmip6), extracting monthly near-surface air
temperature (tas), monthly precipitation (pr), and daily
maximum temperature (tasmax) for 36 globally-distributed cities across
eight independent climate models and four emissions scenarios
(historical, SSP1-2.6, SSP2-4.5, SSP5-8.5). Raw Zarr stores totaled more than
100 GB; we processed them into roughly 165 MB of intermediate CSVs and then
aggregated again to a 9 MB browser-ready dataset. We derived annual extreme
heat metrics (days above 30/35/40 °C, longest annual heatwave, hottest day)
from daily-resolution model runs, computed a climate-analog mapping
(for each city under each future scenario, the historical city whose 12-month
climatology best matches it by RMSE), and pulled global spatial
fields for the world maps. This prototype wires all of that into a
single interactive D3 explorable structured as a top-to-bottom narrative, with
ten distinct chapter-style sections and five of them directly responsive to the
city the reader picks.
The hardest design tension is between honest uncertainty and narrative punch. Eight CMIP6 models disagree by up to 6.6 °C at end of century for a single city under one scenario — that spread is not a bug, it's the real state of knowledge in climate science. But "your city will get between 1 and 8 degrees hotter" is exactly the kind of fuzzy framing that lets readers tune out. Our answer is layered: lead with a confident headline number (the ensemble mean), expose the p10–p90 model spread as a shaded band, and use a pinned scrolly section to walk readers through what each scenario actually looks like rather than leaving them to interpret the spread on their own. The climate-analog framing sidesteps the precision problem entirely — even if we can't pin San Diego's 2080 mean to a tenth of a degree, the answer "it will feel like North Africa" is robust across the ensemble and sticks. The remaining challenge was keeping ten chapters from drowning each other out — the Apple-style chapter layout, with alternating light and dark backgrounds and one big idea per viewport, was our resolution.