TerraScope is an atlas through time. Playback animates data and projections: it does not measure the planet in real time and does not predict a future satellite image.
2100–2200: exploration beyond published series
The hatched part of the timeline is a hypothesis built for TerraScope, not a projection assessed by the IPCC. Global warming is interpolated from the 2081–2100 central value (+1.8 / +2.7 / +4.4 °C) toward the middle of the ranges the IPCC gives for 2300 (AR6, SPM B.5.3, low confidence: +1.0 to 2.2 °C, +2.3 to 4.6 °C, +6.6 to 14.1 °C). The 2071–2099 temperature map is scaled to that warming, over land and relative to 1991–2020.
Desert and dryland fronts keep moving the way they did between 2041–2070 and 2071–2099, in proportion to the extra warming, and only where the front passes: the rest of the published map is kept. Cities continue the growth of their last half-century, never shrinking and never creating new cities. Birth rates and migration stay at the last UN projection (2100), clearly labelled; forests and health series follow the rules below.
Deserts and temperature
The maps of Beck et al. (2023), corrected in January 2026, give the climates of 1931–1960, 1961–1990, 1991–2020, 2041–2070 and 2071–2099. Futures come from the CMIP6 scenarios SSP1-2.6, SSP2-4.5 and SSP5-8.5.
The timeline interpolates between the middle of these periods (1945, 1975, 2005, 2055, 2085), then joins the exploration range. This motion is a visual transition between climatologies, not a yearly measurement. Temperatures show the change in annual means over land relative to 1991–2020; they are not a global anomaly relative to pre-industrial times.
Deserts are the BWh and BWk classes; drylands (steppes) are BSh and BSk. They are shown at 0.2° (majority class of the 0.1° maps) with softened edges; temperatures at 1°. Changes narrower than three cells (about 60 km) relative to 1961–1990 are left out: they are within the uncertainty of the boundary between two climate classes. Colours show the change since 1961–1990: ochre where desert expands, beige where land becomes drier, green where desert retreats (for instance parts of the Sahel). A change of climate class is not a measure of land degradation. Areas are weighted by the true surface of each cell.
Urban growth
GHS-WUP-BUILT-S R2025A — European Commission, JRC (2025). Built-up surface for 1975, 1990, 2005 and 2020, then CRISP projections for 2050 and 2100. The 1 km maps are resampled to 0.1° and interpolated between these dates. The layer is absent before 1975. A cell is drawn as city from about 4% built-up surface; after 2020, areas built since 2020 are tinted amber.
The rendering keeps the geography of the footprints and increases their contrast so they can be seen from space. Building patterns and brightness are a graphic interpretation. CRISP does not depend on the chosen climate pathway.
Forests
From 1960 to 2020: HILDA+ version 2.0 (Winkler et al., 2025), a yearly reconstruction of land use at 1 km (forest classes 40 to 45). After 2020: Chen, Li & Liu (2022), 1 km projections for SSP1-2.6, SSP2-4.5 and SSP5-8.5, applied as a change from their own 2015 map so the two sources join without a jump. Both CC BY 4.0. A 0.2° cell counts as forest when at least half of it is forest. Rust: 1960 forest that is gone. Deep green: forest of that year missing from the satellite image (about 2004), i.e. earlier forest or regrowth.
These scenarios follow the land-use pathways of the IPCC socio-economic models (LUH2): they often project reforestation after 2020 and do not simulate climate-driven Amazon dieback. After 2100, the 2050–2100 trend is extended as an exploration.
Birth rate
UN World Population Prospects 2024, via Our World in Data. Births per 1,000 people per year; estimates up to 2023. For 2024–2100 the rate is computed from annual births and mid-year population in the UN median scenario. It is not the number of children per woman. Country outlines are from Natural Earth, without historical borders; small territories may be missing at this scale.
Migration
UN World Population Prospects 2024: each country's annual net migration (arrivals minus departures), estimated up to 2023 then projected to 2100 (median variant), read every five years and interpolated. The globe colours each country by this balance per 1,000 people; the ranking lists every country by its net number of migrants. A balance describes neither routes nor corridors between countries, and no causal link with climate is implied.
Economic power, debt and military power
Economic power is GDP at purchasing-power parity (2017 dollars). From 1980: the SSP database, 2025 update (OECD ENV-Growth model, IIASA, CC BY 4.0), historical reference then SSP1, SSP2 or SSP5 up to 2100, following the climate pathway you choose (low emissions = SSP1, intermediate = SSP2, high emissions = SSP5). Before 1980, growth comes from the Maddison Project (GDP per capita × UN population, via Our World in Data), chained to the 1980 SSP level. After 2100, the growth rate of the 2090s continues and fades out (30-year time constant).
Public debt is general government gross debt as a share of GDP: IMF World Economic Outlook, estimates then IMF projections to 2031. After 2031 the trend of the last ten years is extended with damping, like the health series (bound 400%): an exploration, not a forecast.
Military power is military spending in constant 2024 dollars: SIPRI Military Expenditure Database via Our World in Data, 1960–2025. No institution projects it. After 2025, TerraScope applies a simple model: each country keeps its 2021–2025 spending as a share of GDP, which then grows with the chosen SSP. Spending is not military capability; wars, alliances and budget choices are not modelled.
Religion
Pew Research Center, The Future of World Religions (2015), via datahub.io (CC BY 4.0): share of Christians, Muslims, religiously unaffiliated, Hindus, Buddhists, folk religions, Jews and other religions in each country, estimated for 2010 and projected for 2020 to 2050 from fertility, age structure, conversion and migration. Pew writes very small shares as "< 1%". The map shows either each country's dominant group (paler when the majority is narrow) or the share of one group; the ranking shows every country's full breakdown. Nothing is shown before 2010. After 2050, each share continues its 2040–2050 change with damping and the shares are rescaled to 100%: an exploration. Pew's 2025 update revised the 2010 and 2020 estimates; these projections still use the 2015 baseline.
Health: obesity, alcohol, cancer
WHO Global Health Observatory, via Our World in Data: share of adults with obesity (BMI ≥ 30, 1980–2024), litres of pure alcohol per person aged 15+ (2000–2020), cancer deaths per 100,000 people (2000–2021). WHO data under CC BY-NC-SA 3.0 IGO. No source projects these series beyond 2050, and no open projection exists. After the last published year, TerraScope therefore extends, as an exploration, each country's slope over its last ten years with damping (30-year time constant: the extra change levels off at about 30 years of slope), and a rise slows as it approaches a plausible bound (obesity 70%, alcohol 20 L, cancer 600 per 100,000). A country whose trend is falling keeps falling, then levels off. These are not forecasts: neither ageing nor health policies are modelled.
Satellite imagery and rendering
NASA Blue Marble, a fixed cloud-free satellite mosaic, with NASA GIBS imagery when you zoom in. WebGL keeps the relief and adds the transitions of land and built-up surfaces. Climate data cannot rebuild the exact vegetation or buildings of a given year. Urban contrast is enhanced and its density quantised for a light display.
Sources accessed on 29–30 September 2026. Beck et al., HILDA+ and Chen et al.: CC BY 4.0; UN statistics via OWID: credited above; GHSL: © European Union, reuse with attribution; outlines: Natural Earth, public domain.
No account needed. Your interactions stay in your browser. The Share button only copies the year, layers, pathway and viewpoint.