Sensors
- VIIRS Day/Night Band (Suomi NPP): 750 m nadir resolution; detects radiances as low as 2 × 10⁻¹¹ W cm⁻² sr⁻¹ in the 0.5–0.9 µm panchromatic band; daily global coverage from a sun-synchronous orbit with a roughly 01:30 local overpass, minimising traffic and industrial noise. Archive from October 2011.
- VIIRS Day/Night Band (NOAA-20): Identical DNB specification to Suomi NPP; launched 2017, providing a second daily overpass and redundancy. The two satellites together reduce single-night cloud-gap probability and allow same-night cross-check of anomalous readings.
- VIIRS Nightfire (Suomi NPP / NOAA-20): Short-wave infrared channels (M10 at 1.61 µm, M11 at 2.25 µm) detect sub-pixel combustion sources such as fires and flares. Useful for separating infrastructure blackout from active burning events that can briefly mask or mimic civilian light loss.
- Sentinel-2 MSI: 10 m optical resolution in visible and near-infrared bands; daytime only. Used to corroborate DNB findings by identifying physical damage to power-generation or transmission infrastructure visible in high-resolution imagery. 5-day revisit at mid-latitudes with both satellites.
What a 750-metre pixel can tell you about a power grid
The VIIRS Day/Night Band was not designed for conflict analysis. It was designed to detect moonlit clouds, fishing boats, and auroral activity. Its sensitivity floor of roughly 2 × 10⁻¹¹ W cm⁻² sr⁻¹ means it can register a single ship at sea or a small town's street lighting from 830 km up. That sensitivity, combined with daily global coverage and a pre-dawn overpass that catches a city before morning traffic and industry add thermal clutter, makes it the most practical open-source instrument for tracking civilian light output over time.
The NOAA/Colorado School of Mines team, working through the Earth Observation Group, developed the methodology for extracting meaningful radiance trends from raw DNB data. Their monthly composites, published openly via the Mines VIIRS Nightfire and VIIRS monthly products, suppress cloud contamination, lunar interference, and ephemeral noise such as fires or flares. The peer-reviewed application of this method to Syria documented a radiance collapse of roughly 83 per cent between 2011 and 2015 in conflict-affected governorates. Later studies applied the same framework to Ukraine following February 2022 and to Gaza.
Monthly composites versus nightly data: two different questions
Monthly composites answer a strategic question: has this city's electrical output changed over weeks or months? They are statistically cleaner, because cloud-free observations are averaged across many nights, and they are directly comparable across years. A governorate that shows a 60 per cent radiance decline over a three-month composite has almost certainly experienced sustained infrastructure disruption, not a single bad-weather night.
Nightly data answers a different question: when did the lights go out? Single-pass DNB imagery, cloud permitting, can resolve an event to within 24 hours. This is relevant for correlating blackouts with specific strikes, grid-switching decisions, or curfew enforcement. The trade-off is noise. A single night's reading can be contaminated by cloud, moonlight, or a large fire. Analysts should treat single-night anomalies as hypotheses, not findings, and confirm them against the next available clear-sky pass.
Three causes of darkness that look identical from orbit
This is where the method's honest limit sits. A radiance drop in DNB data is consistent with at least three distinct causes: physical destruction of generation or transmission infrastructure; deliberate grid shutdown by an operator or authority; and population displacement, where the people who would normally switch on the lights have left. All three produce the same signal. None can be ruled out from DNB alone.
Separating them requires additional evidence layers. Sentinel-2 imagery at 10 m can reveal visible damage to substations, pylons, or generating plant. Displacement can be cross-referenced against UNHCR movement data, mobile-network activity reports where available, or conflict-event databases such as ACLED. Deliberate shutdown leaves infrastructure intact and may show partial restoration on a predictable schedule. The analyst's job is to build a weight-of-evidence argument across these sources, not to treat the DNB number as a verdict.
A further complication is that conflict zones often have active fires, which emit strongly in the short-wave infrared and can bleed into the DNB's broad panchromatic band. The VIIRS Nightfire product, which uses dedicated SWIR channels to characterise combustion sources, helps flag nights where fire contamination may be inflating or masking the apparent civilian light signal.
What the Syria, Ukraine, and Gaza studies established
The Syria study, published by Witmer and colleagues and later extended by the Earth Observation Group, showed that DNB radiance tracked the progression of conflict with enough fidelity to distinguish relatively stable governorates from those experiencing active fighting. The Ukraine analysis, conducted by multiple groups after February 2022, documented targeted blackouts in Kharkiv, Kherson, and Kyiv following strikes on energy infrastructure, with nightly data pinpointing the timing of outages to within a day of reported strike events. Gaza studies applied the same framework to a dense urban environment, where the 750 m pixel is large relative to neighbourhood-scale variation, limiting spatial granularity.
Each study also noted the same caveat: DNB cannot attribute. It can document that radiance fell, approximately when, and approximately where. The cause requires corroboration. This is not a weakness unique to satellite data; it is a feature of any remote sensing method applied to complex human situations. The published literature is explicit about this, which is one reason the method has been accepted in peer-reviewed journals and cited in humanitarian reporting.
Resolution floors and the urban-density problem
At 750 m nadir resolution, a single VIIRS DNB pixel covers an area roughly the size of several city blocks. In a dense city, this means a pixel may integrate light from residential streets, a hospital running on a generator, a military checkpoint with floodlights, and a burning vehicle. The composite methods partially address this by averaging over many nights, but the spatial granularity remains coarse by the standards of optical imagery.
Sentinel-2 does not observe at night, so it cannot directly replace DNB for radiance measurement. Its role is confirmatory: identifying physical damage to infrastructure that would explain a radiance drop, or confirming that buildings are structurally intact in an area where the lights have gone out, which shifts the interpretation toward displacement or deliberate shutdown. The combination of coarse-but-sensitive (DNB) and fine-but-daytime (Sentinel-2) is the standard working pair for this analysis type.
Cloud cover is an irreducible constraint. In winter months at high latitudes, cloud-free DNB observations may be separated by weeks. Monthly composites partially compensate, but in a fast-moving conflict, a three-week cloud gap can obscure a significant event. Analysts should report cloud-gap periods explicitly rather than interpolating across them.
Running this analysis operationally
A practical blackout-monitoring pipeline ingests nightly DNB granules from NASA FIRMS or the NOAA STAR portal, applies a cloud mask derived from the VIIRS cloud-mask product, and computes per-pixel radiance anomalies against a pre-conflict baseline period. Monthly composites are generated in parallel for trend reporting. Alert thresholds, set at a defined percentage drop from baseline across a defined spatial cluster, trigger human review rather than automated reporting, because the ambiguity problem described above makes fully automated attribution unreliable.
Satellize runs this kind of open-constellation analytics pipeline for government clients. The Tonga crop-estimation programme is a different domain, but the underlying approach, systematic radiance time-series against a baseline, is directly transferable to blackout monitoring. Clients who need this capability for a specific area of interest can commission a baseline characterisation, ongoing nightly alerting, and monthly summary reports as separate deliverables, sized to the area and the required latency.
Typical figures
| Spatial resolution (DNB) | 750 m at nadir; degrades toward swath edges |
| Radiometric sensitivity floor (DNB) | ~2 × 10⁻¹¹ W cm⁻² sr⁻¹ (published Earth Observation Group specification) |
| Revisit (Suomi NPP + NOAA-20 combined) | Up to 2 passes per night; daily global coverage per satellite |
| Spectral band (DNB) | Panchromatic 0.5–0.9 µm |
| Typical composite period | Monthly (standard EOG product); custom periods from 7 days upward depending on cloud-free observation density |
| Event-timing resolution (nightly data) | ±24 hours on a cloud-free night; longer gaps during persistent cloud cover |
| Corroborating optical resolution (Sentinel-2 MSI) | 10 m visible/NIR; 5-day revisit at mid-latitudes |
| DNB archive depth | Suomi NPP from October 2011; NOAA-20 from January 2018 |
| Latency (standard open products) | FIRMS near-real-time: ~3 hours; EOG monthly composites: ~2 weeks after month end |
| Delivery formats | GeoTIFF radiance grids, CSV anomaly tables, GIS polygon layers, PDF monthly summary |
Analytics Satellize can run
| Pre-conflict radiance baseline | Median compositing of cloud-masked DNB granules over a user-defined reference period, following Earth Observation Group published methodology | GeoTIFF baseline radiance grid per area of interest, with cloud-observation-count layer |
| Monthly radiance anomaly map | Per-pixel percentage departure from baseline; monthly composites suppress ephemeral noise | GIS polygon layer with radiance-change classes and a PDF summary chart for the reporting period |
| Nightly blackout alert | Single-pass DNB anomaly detection against baseline; cloud-mask applied; threshold set by analyst to control false-alarm rate | Alert feed (GeoJSON or email) within 6 hours of satellite overpass on cloud-free nights |
| Cause-disambiguation assessment | Multi-source weight-of-evidence: DNB radiance trend, Sentinel-2 infrastructure-damage review, VIIRS Nightfire fire-contamination check, open conflict-event database cross-reference | Structured analytical report rating each blackout zone as consistent with physical damage, deliberate shutdown, or displacement, with confidence levels and evidence citations |
| Infrastructure recovery tracking | Rolling monthly composites compared against conflict-period trough; percentage recovery index per administrative unit | Time-series chart and GIS layer updated monthly; suitable for humanitarian or reconstruction-planning reporting |
| Fire-contamination-flagged nightly composite | VIIRS Nightfire SWIR combustion detection used to mask or annotate pixels where active burning may inflate or suppress apparent civilian radiance | Annotated GeoTIFF with fire-affected pixels flagged; separate clean-pixel composite for civilian-light analysis |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.