What Nighttime Lights Tell Us About Cities (And What They Don't)

A Proxy, Not a Measurement

Nighttime lights from the VIIRS Day/Night Band are one of the most intuitive satellite products. Bright areas mean people, dim areas mean fewer people, dark areas mean nobody's home. It's immediately understandable in a way that SAR backscatter or NDVI values aren't.

That intuitive appeal is both the strength and the danger of this data. Nighttime radiance is a proxy for human activity — it correlates with things we care about (economic output, population, infrastructure) but doesn't directly measure any of them.

What the Data Actually Is

NASA's Black Marble product (VNP46A2) provides calibrated nighttime radiance in units of nW·cm⁻²·sr⁻¹. The satellite — Suomi NPP or NOAA-20 — passes over each location at roughly 1:30 AM local time, capturing artificial light emissions.

Key specifications:

• Spatial resolution: ~500m (15 arc-seconds)
• Temporal coverage: Daily composites available from 2012-present
• Processing: Cloud-masked, atmospherically corrected, lunar illumination removed
• Units: Radiance, not a unitless brightness index

That last point matters. The values are physically meaningful — you can compare radiance levels across locations and dates quantitatively, not just visually.

Applications That Work Well

Power Outage Detection

Comparing pre-event and post-event nighttime radiance after a natural disaster or conflict immediately reveals areas that have lost electricity. During Hurricane Maria in Puerto Rico (2017), nighttime lights data showed power restoration progress over weeks and months when ground reports were sparse.

This is one of the most reliable applications because you're looking for an abrupt, localized change — the signal is strong relative to noise.

Urbanization Tracking

Multi-year nighttime lights time series reveal urban expansion patterns. Areas that transition from dark to lit are almost certainly experiencing development. I've found this particularly useful in rapidly growing cities in Southeast Asia and sub-Saharan Africa, where up-to-date land cover maps lag behind actual development.

Conflict and Crisis Monitoring

Dramatic, sustained drops in nighttime radiance can indicate conflict zones, economic collapse, or mass displacement. Syria's nighttime lights decreased by approximately 83% between 2011 and 2015, according to multiple studies. This kind of analysis provides an independent, objective measure that complements (but cannot replace) on-the-ground reporting.

Fishing Fleet Detection

Brightly lit squid fishing vessels are visible in nighttime imagery as isolated bright points on otherwise dark ocean. The patterns are seasonal and regional, and tracking them over time reveals fishing activity trends and potential overfishing areas.

Where It Gets Misleading

Light Pollution ≠ More Activity

A city that adopts LED street lighting may appear brighter without any change in population or economic activity. Conversely, municipalities that reduce unnecessary lighting (dark sky ordinances) will appear dimmer. Neither change reflects what most analysts actually care about.

Seasonal Variation

Higher-latitude cities have longer nights in winter and shorter nights in summer. Snow cover increases reflectance of artificial light. Holiday lighting in December creates temporary brightness peaks. All of these need to be accounted for in time-series analysis.

Saturation in Dense Urban Cores

The VIIRS sensor has a dynamic range limit. The brightest city centers — Times Square, Shibuya Crossing, the Las Vegas Strip — can saturate the detector, making it impossible to distinguish between "very bright" and "extremely bright." For studies focusing on central business districts, this is a real limitation.

Rural Electrification Ambiguity

A newly lit village in rural Africa might be a genuine electrification milestone — or it might be a temporary mining camp, a gas flare, or a large fire. Ground truth or contextual analysis is essential.

Practical Tips

1. Use gap-filled composites rather than single-night observations. Individual nights are affected by clouds, moonlight correction errors, and occasional satellite artifacts.

2. Compare the same month across years to minimize seasonal effects.

3. Cross-reference with other data. Nighttime lights plus daytime optical imagery (from Sentinel-2) gives you a much more complete picture than either alone.

4. Be cautious with thresholds. There's no universal radiance value that separates "urban" from "rural." What counts as bright depends entirely on regional context.

5. Watch for gas flares. Oil and gas flaring can produce extremely bright, persistent light sources that have nothing to do with human settlement. The Niger Delta and parts of Siberia are common problem areas.

Quantitative Reference: What the Numbers Mean

Working with VIIRS DNB data is easier once you have a feel for typical radiance values. These are approximate nW·cm⁻²·sr⁻¹ values from the VNP46A2 Black Marble monthly composite:

Context	Typical Radiance (nW·cm⁻²·sr⁻¹)
Pristine dark sky (remote wilderness)	< 0.2
Rural agricultural area (sparse settlement)	0.5–2
Small town or suburban periphery	2–10
Mid-sized city (300k–1M population)	10–50
Dense urban core of a major city	50–200
Saturated detector (city centers, major ports)	> 200 (clipped to max)
Gas flare in oil field	500–3,000 (bright point source)

These ranges shift with latitude — longer winter nights increase composite values at high latitudes — and vary with local lighting practices. The transition from sodium vapor street lights to LEDs, which has occurred across much of Europe and North America since 2015, shifted spectral output in ways the DNB responds to differently, causing modest apparent radiance decreases in converted cities even as total light pollution increased.

Power outage detection threshold: A pixel that drops more than 50% from its seasonal average in a monthly composite and remains low for two or more consecutive months almost certainly reflects a real disruption. Single-month dips are often cloud mask imperfections or unusual atmospheric conditions rather than genuine outages. The persistence criterion is key — weather effects rarely last more than a few days, but infrastructure disruption typically lasts weeks to months.

A note on gas flares: If you're monitoring an area in or near petroleum-producing regions (Nigerian Delta, Siberian lowlands, Permian Basin), filter flare pixels before any analysis — they can be orders of magnitude brighter than nearby settlements and will dominate spatial statistics if not removed.

Try It

Off-Nadir Delta provides access to VIIRS nighttime radiance data. Load a familiar city and explore the patterns — the core should be brightest, with radiance decreasing outward into suburbs, then dropping off sharply at the urban edge. That gradient itself tells you something about the city's spatial structure.