Each city measured against its own lit footprint and its own history, from VIIRS satellite passes. Never against other cities, because that would only map wealth.
Where cities go dark
Circle size is how bright the city normally is. Click to pin, click again to unpin.
Less darkMore dark
City ranking
Click a bar to pin that city into the comparison.
Over time
Pin up to five cities to compare. Nothing pinned shows the filtered average.
Seasonality
Average by calendar month across all years. The rainy season shows up as cloud, which is a caveat, not a blackout.
Does city size explain it?
Normal brightness against share going dark. If the cloud has no slope, dimming is not simply a size or wealth effect.
By country
Cities averaged within each country. Bar width shows the range across its cities.
All cities in view
Click a column heading to sort. Click a city name to pin it.
Read this before drawing conclusions. Nothing here is a finding yet.
Observations are sampled near new moon so lunar brightening is controlled, and nights with poor coverage are discarded. But thin cloud is not corrected for, and thin cloud dims a city in exactly the way a partial outage does. A single low night is more likely weather than a blackout. Read the ranking and the trend, not any individual number.
Known bias. Generators and rooftop solar dim less during an outage and they track wealth, so this systematically under-detects outages in richer areas and richer countries.
What the number means. The share of a city's normally lit area sitting below half its own usual brightness. Because "usual" is the 90th percentile of that city's own nights, a typical night sits below it by construction, so the absolute level runs high everywhere. The comparison between cities is the informative part, not the level.