Satellite pictures of the planet's lights, every year from 2012 to 2024, with a modelled look back to 2000. Where the world lit up, where it went dark, and why a darker picture from space does not always mean a darker street.
From orbit, at half past one in the morning, the inhabited world draws itself in light. Coastlines glow, roads string cities together like beads, and whole countries show up as dark gaps. Since 2012 an instrument called VIIRS, on a weather satellite that passes over every place at about the same hour each night, has measured that light with enough care to compare one year with the next. I wanted to use those measurements to answer three plain questions. How much brighter has the planet become? Where did the new light come from? And where did the lights go out?
The yearly pictures come from the Annual VIIRS Night Lights of the Earth Observation Group at the Colorado School of Mines, prepared as one consistent stack of world maps by Tomislav Hengl of OpenGeoHub. I used them at about 15 km and 7.4 km resolution and added up the light over every country and around every city of half a million people or more. To keep the measure simple, a square kilometre shining at 1 nanowatt per square centimetre per steradian (the unit the satellite measures in; a bright city centre measures several tens) counts as one light unit. Only the years 2012 to 2024 are observations. Hengl also extended the series back to 2000 with a statistical model; I show those years on the map, clearly marked, but every number below comes from the observed years.
Add up every lit square kilometre and the Earth gave off 57% more light in 2024 than in 2012: 52.3 million light units against 33.2 million. Single years are noisy, so I also compared the average of the first three observed years with the last three, which gives a rise of 41%. Either way the direction is not in doubt: the total fell from one year to the next only in 2014, 2015 and 2020. The lit area grew even faster than the light. Land that averaged at least 1 nW across a 15 km cell covered 9.3 million km² in 2012 and 15.1 million km² in 2024, up 61%, from 6.8% to 10.9% of the land.
Light seen from space each year
Total light on the 15 km grid, millions of light units. Solid: observed by VIIRS. Dashed: modelled backwards from the observations, a sketch only. Hover for values.
The modelled years tell a gentler story, from 21.6 million units in 2000 to 24.9 million in 2011, and they do not meet the observations: the first observed year is 34% brighter than the last modelled one. That seam is a reminder of what the early years are. They are a curve fitted to each place's observed light and run backwards, good at showing which places were already lit, but not a record of what anyone saw.
The light is shifting east and south. North America gave off 24.2% of the world's light in 2012 and 18.7% in 2024; Europe and Central Asia went from 26.6% to 22.7%. East Asia and the Pacific rose from 14.7% to 18.1%, and South Asia from 4.5% to 7.4%, the fastest growth of any region (159%). Sub-Saharan Africa's light grew by 70%, but from so low a start that its share barely moved: 2.6% of the world's light for 14% of its people.
Each region's share of the world's light, 2012 and 2024
World Bank regions as in Natural Earth. Open circle: 2012. Filled: 2024. Grey tick: share of the world's people. Hover for the change in light.
Most of the new light is in Asia. China's light more than doubled (2.1 times) and India's grew 2.7 times; between them they gave 26% of all the extra light the planet gave off in 2024 compared with 2012. China's share of the world's light went from 7.9% to 10.5%, India's from 3.9% to 6.6%, while the United States, still the brightest country, went from 21.6% to 16.9%. Of the 141 countries with at least a million people and some light to start from, 47 at least doubled their light and 16 at least tripled it.
The fastest risers started from very little. Nepal's light grew 9.3 times; in 2012 the country was living with daily power cuts of many hours, which were ended over 2016 to 2018 as new supply arrived, and Kathmandu's light grew 4.9 times. Then come Cambodia (7.0×), Ghana (6.4×), Tanzania (6.0×), Côte d'Ivoire (5.5×), Senegal (5.2×), Kenya (4.9×) and Ethiopia (4.2×). Africa's cities stand out among the big cities too: Kumasi's light grew 4.4 times, Addis Ababa's 3.6, Accra's 3.3, Dar es Salaam's 3.0 and Abidjan's 3.0. Night light follows electricity and economic activity loosely, so these are signs of new grids, roads, homes and businesses, but they are not a measure of any one of them.
Brightened and dimmed: the biggest changes, 2012 to 2024
Change in each country's total light, on a doubling scale (each step is twice or half). The fifteen fastest risers and every country that lost a tenth or more of its light; countries with at least a million people and 2,000 light units in 2012. Hover for the three-year comparison.
Some rises need a warning label. Iraq's light grew 3.2 times and Iran's 2.4 times, and part of that is not cities at all but the flames of gas flares burning off unwanted gas at oil fields, which the satellite records as light. Basra, among the oil fields of southern Iraq, grew 3.5 times. Yet Baghdad's own light grew 3.9 times, and Mosul's light fell by 70% between 2013 and 2015, after the city was taken by the Islamic State group in 2014, then recovered to 4.1 times its 2012 level by 2024.
Only 17 of the 141 countries gave off less light in 2024 than in 2012, and 12 of them by a tenth or more. Three quite different things lie behind the dimming, and the satellite cannot tell them apart. Turn on "Lit up or went dark" on the map to see them: gold where light grew, magenta where it fell.
The steepest falls are wars. Ukraine's light fell by 73% between 2021 and 2022, the year of the full-scale Russian invasion and the strikes on its power grid, and in 2024 it was still 53% below 2012 (Crimea is counted with Russia here, following Natural Earth's de facto borders). Kharkiv lost 98% of its light, Zaporizhzhia 96% and Odesa 72% from their peaks. Syria, already at war in 2012, lost a further 65% of its light by 2015; Aleppo lost 97%. Yemen's light fell 68% between 2012 and 2016, as its war escalated in 2015, and Sanaa's 95%. In Sudan, Khartoum lost 77% of its light between 2022 and 2024, after fighting broke out in the capital in 2023.
Cities that went dark, and cities that lit up
Light within 20 km of each city, every year on the 15 km grid (observed years only), scaled to each city's own peak. Click a city to find it on the map.
Some countries dimmed without a war. Lebanon's light fell 71% between 2019 and 2022, as an economic collapse left the state power company short of fuel; Beirut lost 73% of its light before part of it came back. Venezuela's light fell 41% between 2014 and 2020, through an economic crisis and the nationwide blackouts of 2019 (Maracaibo lost 53%), then rose again. Today about half of the country's measured light (53% in 2024) comes from the oil fields of Monagas, where gas is flared, which helps explain why its 2024 total is higher than 2012's. Afghanistan's light fell 52% between 2013 and 2022, a fall that deepened after the Taliban returned to power in 2021.
The third kind of dimming is the most surprising, because it is happening in rich, peaceful countries. France gave off 40% less light in 2024 than in 2012, the Netherlands 26% less, Switzerland 19% less and the United Kingdom 13% less; Paris and London each lost about a quarter of their light. Nothing like that happened on the ground. Two changes explain most of it. First, towns have been replacing orange sodium street lamps with white LEDs, and the satellite's sensor cannot see blue light, so the same street lit with LEDs looks dimmer from space even when it is as bright or brighter to the eye. Second, the satellite passes at about half past one in the morning, and more and more towns, thousands of French communes among them, now switch street lights off for part of the night; France's energy-saving rules of 2022 also put out illuminated advertising between one and six in the morning. In the Netherlands, the drop around The Hague (60%) and Rotterdam (49%) is mostly the great greenhouses of the Westland, whose growers cut their lighting when gas prices soared in 2022 and have been moving to LEDs.
So a fall in measured light can mean a city under bombardment, a country that cannot pay for fuel, or a town that has changed its lamps. People who count the stars they can see from the ground report that the night sky has been getting brighter much faster than the satellites suggest, largely because of that blind spot for blue light. Gas flares add light that has nothing to do with people's homes and can make an oil country look brighter or darker as flaring rises and falls; Gabon's 23% fall may be partly that.
Divide each country's light by its people (Natural Earth's population estimates, mostly for 2019) and set the world average to 100. Qatar scores 1,302, Libya 963 and Saudi Arabia 837: small populations, a great deal of lit infrastructure and, in several of them, gas flares. The United States scores 396, Russia 433, the United Kingdom 85, China 58 and India 37. At the other end, Ethiopia scores 3.0, Uganda 5.0, DR Congo 5.5 and Bangladesh 5.8. Sub-Saharan Africa as a whole scores 18, South Asia 31 and North America 392. That is a gap of more than 21 to one between the average North American and the average person in sub-Saharan Africa, in light visible from space.
Light per person, 2024 (world average = 100)
The ten countries with most and least light per person, among those with at least a million people; log scale. Hover for values.
| Country | 2012 | 2024 | Change | 3-year | Per person |
|---|
Light units, thousands (one unit is a square kilometre shining at 1 nW/cm²/sr). 3-year: change between the averages of 2012 to 2014 and 2022 to 2024. Per person: world average = 100. Download all of it as data/countries.csv.
Before VIIRS, night lights came from the older DMSP satellites, flown since the 1970s, whose cameras were coarser and saturated in city centres. Researchers have stitched DMSP and VIIRS into a single DMSP-style series from 1992. The extract I used holds only one picture from it, the average light of every square kilometre over all the years from 1992 to 2024, so it cannot show change, but it is the best map of where light has been for a generation. Choose "Long-term average" on the map to see it; it matches the VIIRS map closely (correlation 0.70 on a log scale).
Night lights. Hengl's yearly layers are the Earth Observation Group's Annual VIIRS Night Lights V2, version 2.1, at 500 m, stored in tenths of a nW/cm²/sr. I used the overview levels built into his files: about 15 km for every year and 7.4 km for 2012 and 2024. In this release the layers for 2022, 2023 and 2024 are stored in whole nW rather than tenths (on bright cells, 2022 divided by 2021 has a median of 0.100), so I converted every year to whole nW and rounded it the same way, set negative values (noise) to zero, and left out the strip north of 75.1° N, which has no observations. Light units are radiance times area. About 1.9% of the light falls outside any country (boats, oil platforms and coastal glare) and is counted in the world total but not in any country's.
Countries, cities and people. Each 15 km cell belongs to the Natural Earth country that holds most of its land; Natural Earth's de facto borders put Crimea in Russia and do not imply any view on disputed places. Cities are Natural Earth's populated places with at least 500,000 people, measured within 20 km of the city's point, so neighbouring cities can overlap. Light per person uses Natural Earth's population estimates, mostly for 2019. Rankings leave out countries with fewer than a million people or under 2,000 light units in 2012, and cities with under a million people or very little light in 2012, because small numbers make huge ratios.
What the light is. The satellite passes at about 1.30 am local time. It sees streets, homes, roads, ports, industry, greenhouses and gas flares; the product is designed to leave out fires, moonlight and other passing light. It does not see blue light below about 500 nm, so changing from sodium lamps to white LEDs makes places look dimmer from space. Part-night lighting, clouds, snow and the angle of view add noise, which is why I check single years against three-year averages: on that check 6 of the 141 countries change direction, every one of them within a tenth of no change on one of the two measures.
The modelled years. The years 2000 to 2011 were made by Hengl by fitting a logistic curve to each pixel's observed light from 2012 and running it backwards. They are shown for orientation only, they do not join smoothly onto the observations, and no claim on this page rests on them. The DMSP layer is a long-term average, not a yearly series. A DMSP standard deviation layer in the same extract could not be read because the file was damaged, so it is not used.