Dying Pale Blue Dot
Your footprint

Methodology & sources

Every number, traced to its source

The whole point of this site is that the story is real. So nothing here is invented or projected-as-observed: every figure traces to a cited primary source, missing values render as “—” (never 0), and the same numbers are exposed as an open API you can check yourself.

Back to the overview

A warming world, in our own hand

Climate deep dive →

Temperature is shown as an anomaly — the departure from the 1951–1980 average — because anomalies combine consistently across thousands of land and ocean stations. NASA’s GISTEMP v4 blends station records with sea-surface measurements; the value for the current year is a year-to-date mean and firms up as more months report.

Atmospheric CO₂ is the annual mean from NOAA’s Mauna Loa observatory, the longest continuous direct record (the Keeling Curve, since 1958). Fossil CO₂ emissions are the Global Carbon Project’s estimate of CO₂ from fossil fuels and industry, excluding land-use change, distributed via Our World in Data.

The 800,000-year context chart is the published Antarctic ice-core composite (Bereiter et al. 2015, NOAA NCEI study 17975): air bubbles trapped in ice preserve samples of the actual ancient atmosphere. Every plotted point is a real published sample — the record is thinned for display by dropping points, never by averaging or interpolating — and the years after the ice record ends (2001, Law Dome) continue with the same live Mauna Loa series charted above, exactly as NOAA’s file header directs.

Greenhouse gases (fossil & industry) is the broader companion to the fossil-CO₂ line: it adds methane, nitrous oxide and fluorinated gases from energy and industry, all converted to a common 100-year CO₂-equivalent (Jones et al. 2025, via OWID). It excludes land-use change and so is a conservative floor on total warming pollution — already over 43 Gt CO₂e a year.

Two views of the cause: CO₂ per person (Global Carbon Project) is the global average fossil emission per head — it has roughly doubled since 1950 and, tellingly, has barely fallen since, even as efficiency improved. Fossil fuel production (Energy Institute) is the coal, oil and gas we extract each year in petawatt-hours; despite every climate pledge, it has kept climbing decade on decade.

Forest area (FAO, via OWID) is the world’s total forest cover; it has fallen by about a Mexico-sized area since 1990. It is the slow-moving net stock behind the faster `forest loss` series above — the standing forest that is still, on balance, shrinking.

Atmospheric methane (CH₄) is NOAA’s globally-averaged annual mean (parts per billion) from a network of marine surface sites. Methane is a far more potent greenhouse gas than CO₂ — roughly 80× over 20 years — though it is shorter-lived; it has risen from ~1,645 ppb in 1984 to over 1,930, with a sharp acceleration since 2007.

Nitrous oxide (N₂O) — the third major greenhouse gas, ~270× as warming as CO₂ and very long-lived — is NOAA’s global annual mean (the global record begins in 2001). It has climbed steadily from ~316 to ~339 ppb, driven mainly by nitrogen fertiliser and manure in agriculture.

Tree-cover loss (Global Forest Watch / University of Maryland) counts the annual removal of tree canopy from any cause — logging, fire, clearing for farmland — at ≥30 m resolution. It is loss, not net change, so it does not subtract regrowth. This series is currently seed-only: no faithful public year-by-year feed is wired for live refresh (see docs/DATA_SOURCES.md).

And two hopeful counter-trends. Renewables’ share of global electricity (Our World in Data, from Ember + the Energy Institute) stalled near 18% through the fossil-heavy 1990s–2000s, then climbed past a third as wind and solar scaled. And the electric-car share of new sales (IEA, via OWID) has gone from almost nothing to a quarter of all new cars in barely a decade — the clearest signs that the curves CAN bend the right way.

Rising seas, vanishing ice

Oceans & Ice deep dive →

Global mean sea level here is the Church & White (CSIRO) tide-gauge reconstruction, anchored near zero around 1900. It is a reconstruction, not a single instrument, so early decades carry wider uncertainty; satellite altimetry since 1993 shows the rise has accelerated. This series is seed-only — no faithful public live feed is wired yet.

Ocean heat content is NOAA NCEI’s estimate (Levitus et al.) of the heat stored in the upper 700 m of the ocean, in units of 10²² joules relative to the 1955–2006 mean — the clearest single measure of global warming, since the ocean absorbs over 90% of the excess heat. The numbers are abstract but the trend is not: it has climbed from roughly −3 to +23 since 1955.

Sea surface temperature is the Met Office Hadley Centre’s HadSST anomaly — the temperature of the ocean SKIN, vs the 1961–1990 mean. It is a different thing from the heat content above (surface vs the whole upper ocean) and from the land+ocean global anomaly on the Climate page; here it isolates the surface, now warmer (~+1.1 °C) than at any point in the instrumental record. The current year is preliminary.

Arctic sea ice is the September monthly-mean extent from the NSIDC Sea Ice Index (passive-microwave satellites, continuous since 1979). September is the annual minimum — the clearest measure of how much multi-year ice survives the melt season. (This is floating SEA ice; melting it does not itself raise sea level.)

Antarctic sea ice is the annual MINIMUM (February) extent from the same NSIDC record (via OWID). For decades it drifted sideways or even grew slightly, confounding expectations — until 2023, when the summer minimum collapsed to the lowest in the satellite era and has stayed near those lows since. Like the Arctic, this is floating sea ice; its loss removes a bright, reflective surface and exposes dark, heat-absorbing ocean.

Mountain glacier mass is the cumulative mass balance of the WGMS reference glaciers — the ~40 best-monitored glaciers worldwide with 30+ unbroken years of direct field measurement — in metres of water equivalent, served via the Met Office Climate Dashboard. Around 1970 the running balance crossed zero; by 2025 it had fallen to about −28 m w.e., and the recent slope is the steepest in the record. Mountain glaciers matter beyond sea level: roughly two billion people depend on glacier-fed rivers for water.

The Greenland and Antarctica ice-sheet masses are from NASA’s GRACE and GRACE-FO satellites (via OWID), which weigh the ice by sensing its gravity. We show the cumulative change in gigatonnes since 2002 — about −4,900 Gt for Greenland and −2,700 Gt for Antarctica by 2020. Unlike sea ice, this is LAND ice: every tonne that melts adds directly to sea-level rise, and Antarctica holds by far the most.

Ocean pH is surface seawater measured at Station ALOHA in the North Pacific (Hawaii Ocean Time-series, via OWID) — one of the longest open-ocean chemistry records. As the sea absorbs CO₂ it forms carbonic acid; pH has fallen from ~8.11 to ~8.04 since the late 1980s (a pH scale is logarithmic, so that is a ~18% rise in acidity). It is one station, not a global mean, but it tracks the basin-wide trend closely.

All of these are downstream of emissions — warming for sea level and ice, dissolved CO₂ for acidity — so the actions that bend them are the same ones that bend the climate curves.

The web of life, unravelling

Life deep dive →

The Living Planet Index (WWF & the Zoological Society of London, 2024 report, via OWID) aggregates the relative abundance of ~35,000 monitored populations of ~5,500 vertebrate species — mammals, birds, fish, reptiles, amphibians — indexed to 1970 = 100. The 2020 value of ~27 means those populations have fallen by ~73% on average.

It is an INDEX of average change, not a count of animals or species, and not a measure of extinction. A few steep regional declines (notably freshwater and the tropics) pull the global average down, so read it as the direction and scale of pressure on wild populations, not a headcount.

The Red List Index (IUCN & BirdLife, UN SDG 15.5.1, via OWID) is the complement to the Living Planet Index: where the LPI tracks population abundance, the RLI tracks EXTINCTION RISK. A value of 1 means all assessed species are Least Concern; 0 means all are extinct. The world index has fallen steadily — every year, more species slide toward the threatened categories faster than any recover.

In the ocean, overfishing tells the same story: the FAO’s share of assessed marine stocks fished beyond sustainable limits (via OWID) has climbed from ~10% in 1974 to over a third today — a steady erosion of the seas’ ability to feed both wildlife and us.

Protected land (UNEP-WCMC / World Bank, via OWID) is the share of the world’s land inside protected areas — now about a sixth, and slowly rising toward the global “30 by 30” goal. It is a genuine bright spot, with one honest caveat: a line on a map is not the same as effective protection, and it says nothing about the oceans or how well habitats actually fare inside.

The drivers of the decline are land- and sea-use change (habitat loss), overexploitation, pollution, invasive species and — increasingly — climate change. That is why eating less meat, cutting emissions, and refusing throwaway plastic all show up below as ways to ease the pressure.

A throwaway world

Waste & Plastic deep dive →

Plastic production is global annual output of primary plastics, compiled by Geyer et al. (2017) and extended by the OECD, distributed via Our World in Data. It counts polymer resin produced, not waste generated.

Material footprint per person (UN International Resource Panel, SDG 12.2.1, via OWID) is the total raw material — biomass, fossil fuels, metals and minerals — extracted worldwide to satisfy each person’s consumption, wherever in the world that extraction happens. It is the upstream measure of throughput behind all the waste below: it has risen from about 9 to 12 tonnes per person since 2000.

Fertiliser use (FAO, via OWID) is the global tonnage of synthetic nitrogen, phosphate and potash applied to farmland each year. It has multiplied roughly sixfold since 1961. The nutrients that boost crop yields also run off into rivers and seas, where they feed algal blooms and create oxygen-starved “dead zones” — the planetary boundary scientists call disrupted biogeochemical flows.

Freshwater use (IGB long-run reconstruction, via OWID) is the world’s total annual withdrawal of freshwater for farms, factories and homes, in cubic kilometres, reconstructed back to 1901. It has risen roughly sixfold over the century — overwhelmingly to irrigate crops — draining rivers, lakes and aquifers faster than they refill. The series ends in 2014 (the latest in this long-run compilation) and recent decades show a slowing, plateau-like trend.

The end-of-life breakdown is the OECD Global Plastics Outlook’s 2019 baseline: the share of plastic waste that is landfilled, mismanaged, incinerated or recycled, plus the estimated ~1.7 Mt/yr that reaches the ocean.

Municipal solid waste per person is from the World Bank’s What a Waste 2.0 (2018), grouped by national income. The gap between low- and high-income generation is the clearest signal that waste tracks consumption, not population.

The air billions of us breathe

Air deep dive →

These are population-weighted annual mean PM2.5 concentrations by country from the IQAir 2023 World Air Quality Report, aggregated from ground monitors and validated low-cost sensors. PM2.5 — particles under 2.5 microns — lodges deep in the lungs and crosses into the bloodstream.

The reference line is the WHO 2021 air-quality guideline of 5 µg/m³ annual mean; bars above 35 µg/m³ (seven times the guideline) are drawn in alarm red. This is a single recent snapshot, not a time series, and is currently seed-only.

Most PM2.5 comes from combustion — transport, coal and biomass burning — so the same shift away from burning fossil fuels that cools the climate also clears the air.

Average PM2.5 exposure (IHME Global Burden of Disease 2021, via OWID) is the population-weighted global mean concentration each person breathes, the long-run companion to the country snapshot above. It climbed to a peak around 2013, then fell as China and others cleaned up — but at ~31 µg/m³ it is still roughly six times the WHO safe limit of 5. Atmospheric aerosol loading is the ninth planetary boundary, and this is the clearest global measure of it for human health.

One atmospheric problem we did fix: ozone-depleting substances. The index here (UNEP Ozone Secretariat, via OWID) tracks global consumption of controlled substances — CFCs and the like — relative to their 1986 level. After the 1987 Montreal Protocol, consumption fell by about 99%, and the stratospheric ozone layer is now slowly healing. It is the clearest proof that a co-ordinated global response to an atmospheric threat can work.

The Antarctic ozone hole panel shows the effect side of that story: the satellite-observed annual maximum extent (NASA Ozone Watch, via OWID). It carries no good/bad badge, and is excluded from the state-of-planet tally, because year-to-year size is driven by stratospheric weather — 2020–23 were large for meteorological reasons — so a short-run trend misleads in both directions. The honest recovery signal is the falling consumption curve above plus the WMO/UNEP 2022 assessment, which expects Antarctic ozone back at 1980 levels around 2066.

Synthesized & derived surfaces

A few surfaces don’t plot a single measured series — they synthesize the cited data (a tally, a countdown, a projection) or report a cited assessment. None are observed data, so each is fenced off: computed at render time, never written to the database or served by the API, and visibly labelled. Here is every one, with its source.

Check it yourself

The same cited numbers are served as a read-only JSON API with full provenance. Where a faithful live feed exists, a daily job refreshes the series; otherwise the cited static baseline stands, and a failed fetch never overwrites good data.

Open data API →