What we know, how we know it, and where we're heading. A 13-slide synthesis of paleoclimate records, attribution science, and the road ahead.
Sunlight arrives mostly as visible light. Earth re-radiates heat as long-wave infrared. CO₂, CH₄, water vapor, and N₂O have molecular bond resonances that absorb that IR — and re-emit it in all directions, including back down.
We don't need a thermometer. Every glacial cycle is preserved in air bubbles, isotopes, and growth rings — independent archives that converge on the same story.
EPICA Dome C (Antarctica) drilled to bedrock — 800 ka of trapped air. Vostok confirms. δ¹⁸O gives temperature; bubbles give CO₂ directly.
Dendrochronology resolves year-by-year temperature back ~12,000 years for some regions. Width and density encode growing-season warmth.
Foraminifera shells in seafloor cores extend the record to tens of millions of years. Mg/Ca and δ¹⁸O reconstruct deep ocean T.
| Archive | Reach | Resolution | Proxy |
|---|---|---|---|
| Ice cores | 800 ka | annual–decadal | δD, δ¹⁸O, CO₂ |
| Tree rings | ~12 ka | annual | width, density |
| Speleothems | ~500 ka | annual–decadal | δ¹⁸O, U/Th dating |
| Foraminifera | ~70 Ma | centennial+ | δ¹⁸O, Mg/Ca |
Reconstructions disagree on tenths of a degree before 1850. They all agree on what came after: a sharp, sustained departure from any prior 2,000-year baseline.
Charles David Keeling started measuring atmospheric CO₂ in 1958. Same instrument site, same calibration. The curve has not stopped rising for a single year.
No paleoclimate record shows CO₂ rising this fast. Even at the PETM (56 Mya), the increase took thousands of years.
A warming Sun would heat the whole atmosphere. Greenhouse warming has a specific signature: it warms the lower atmosphere while cooling the stratosphere. That's exactly what we observe.
The stratosphere has cooled ~1°C since 1980. A brighter Sun would warm it. Only enhanced GHG greenhouse forcing produces this pattern — and the magnitude matches models.
Arctic warming is ~4× global average. Predicted by physics (ice-albedo feedback, Planck term) decades before measurement. Antarctica lags due to deep ocean heat sink.
Fossil-fuel CO₂ is depleted in ¹⁴C (it's ancient) and ¹³C. Atmospheric ¹³C/¹²C ratio has dropped exactly as expected. The new CO₂ is provably from coal, oil, gas.
If the Sun were the cause, days would warm more than nights. They don't. Nights warm faster because GHG insulation acts strongest when there's no incoming sunlight.
"Climate sensitivity" is the field's central number: how much does Earth warm at equilibrium for each doubling of atmospheric CO₂? After 50 years of refinement across paleo, models, and observations, the answer has stayed remarkably stable.
| Scenario | 2100 CO₂ | ΔT |
|---|---|---|
| SSP1-1.9 (best) | ~400 ppm | +1.4°C |
| SSP2-4.5 (mid) | ~600 ppm | +2.7°C |
| SSP3-7.0 | ~800 ppm | +3.6°C |
| SSP5-8.5 (worst) | ~1100 ppm | +4.4°C |
Most warming is gradual and reversible. Some is not. Several Earth subsystems can flip into a new state and stay there for centuries — millennia — even if we eventually pull CO₂ back down.
Threshold: ~1.5–3°C. Once crossed, melt becomes self-sustaining via altitude feedback. Commit: ~7m sea level rise.
Marine ice-sheet instability may already be triggered for parts of WAIS. Commit: ~3m sea level over centuries.
Atlantic overturning has slowed ~15% since 1950. Full collapse would freeze NW Europe, shift monsoons, raise NE-US sea level.
~17% deforested, ~3.5°C local warming → savannification cascade. Forest becomes net carbon source.
~1,500 Gt carbon frozen — 2× atmosphere. Warming releases CO₂ + CH₄ as a self-amplifying loop.
Northern shift; insect outbreaks; megafires. Carbon sink → carbon source already in some regions.
~99% loss expected at 2°C. Mostly gone already at 1.5°C. Functional collapse this century.
West African and South Asian systems sensitive to aerosol/GHG balance. Shifts disrupt billions.
A warmer atmosphere holds more water (~7% per °C — Clausius-Clapeyron). Wets gets wetter, dries get drier, and the swing between them sharpens. Heat waves that used to occur once a decade now occur three to five times.
In 2010, solar PV cost $0.30/kWh. In 2026, it's under $0.03/kWh — a 90% drop, ahead of every forecast. Battery storage followed. Wind too. The remaining problem is deployment speed and grid integration, not invention.
| Technology | 2010 LCOE | 2026 LCOE | Δ |
|---|---|---|---|
| Utility solar PV | $0.378/kWh | $0.029/kWh | −92% |
| Onshore wind | $0.099/kWh | $0.033/kWh | −67% |
| Offshore wind | $0.188/kWh | $0.075/kWh | −60% |
| Li-ion storage | $1191/kWh pack | $95/kWh pack | −92% |
| EV battery | ~$1100/kWh | ~$110/kWh | −90% |
Heat, transport, industry. ~50% of fossil use disappears as electricity decarbonizes.
Cement, steel, aviation, shipping — need H₂, biofuels, CCS, or substitutes. Solvable, expensive.
DAC + bio + enhanced weathering. Currently ~Mt scale; needs Gt by 2050. Expensive insurance, not a substitute.
Even on the most aggressive mitigation path, ~1.5°C is essentially baked in by inertia. The question for cities, water systems, agriculture, and insurers is no longer "if" but "how fast."
Cool roofs, urban canopy, permeable surfaces, district cooling. Phoenix, Singapore, Medellín are showing the playbook. Heat-island effect can add another +5°C locally.
Storage diversification, leak reduction, desalination where viable, water-trading markets. Cape Town's Day Zero a preview, not an exception.
Drought-tolerant cultivars, shifting plant zones north + uphill, regenerative practices, irrigation efficiency. Yield gaps still large in many regions.
Markets retreating from FL, CA wildfire zones, Gulf Coast. Repricing forces relocation faster than policy. Watch this space — it's the leading indicator.
Adaptation does not substitute for mitigation. It buys time and reduces suffering at any given level of warming — but the cost rises non-linearly with each fraction of a degree.
Pretending 1.5°C is still alive doesn't help. The carbon budget at 50% probability was ~500 GtCO₂ in 2020. We've spent over half. At ~40 Gt/yr emissions, the math is unforgiving.
Without dramatic acceleration, we'll cross 1.5°C in the early 2030s. Overshoot is now the planning baseline for most credible scenarios. Some impacts (coral, low atolls, glaciers) are committed.
Emissions per dollar of GDP are falling. China's emissions may have peaked. Coal is in structural decline. EVs are eating ICE share. Renewables now dominate new generation. The slope is bending — just not fast enough yet.
| Outcome | 2100 ΔT | Plausibility 2026 |
|---|---|---|
| 1.5°C, no overshoot | +1.5°C | very unlikely |
| 1.5°C with overshoot | +1.6–1.8°C peak | plausible w/ CDR |
| 2°C | +1.9–2.1°C | defensible |
| Current policies | +2.5–3.0°C | tracking here |
The science is mature, the data is open, and the explanations are abundant. Below: primary sources, syntheses, and accessible explainers.
"The stone age didn't end because we ran out of stones." — Sheikh Yamani.
The fossil age won't end because we run out of fossils, either.