VIIRS day-night band nighttime light monitoring
The VIIRS Day-Night Band detects low-radiance visible and near-infrared emissions at 750 m nadir resolution, enabling electrification mapping, economic proxies, gas-flare monitoring and conflict damage assessment. Saturation, lunar contamination and a broken heritage link to DMSP-OLS are the limits every analyst must account for.
Sensors
- VIIRS DNB (Suomi-NPP): 750 m nadir resolution, single panchromatic band 500–900 nm, daily global coverage, on-orbit since October 2011. The DNB's three gain stages give it a dynamic range of roughly seven orders of magnitude, allowing detection of radiances from moonlit snowfields down to faint fishing-vessel clusters.
- VIIRS DNB (NOAA-20): Identical sensor design to Suomi-NPP DNB, launched January 2018. Flying in a coordinated orbit roughly 50 minutes apart from Suomi-NPP, the two satellites together improve temporal sampling and provide cross-calibration redundancy. NOAA-21 (launched 2022) carries a third DNB.
- DMSP-OLS (heritage archive): Operational from the late 1970s through the 2010s at roughly 2.7 km resolution with no on-board calibration and significant inter-satellite gain differences. The archive extends the nighttime-light record back decades but cannot be directly compared to VIIRS DNB radiance values without intercalibration. Treat it as a qualitative trend record, not a radiometric one.
- VIIRS Nightfire (NOAA/Mines): A specific VIIRS product that uses the DNB alongside the M-band shortwave infrared channels to detect and characterise sub-pixel combustion sources including gas flares and biomass burning. Published by the Colorado School of Mines Earth Observation Group under NOAA contract.
What the Day-Night Band actually measures
The VIIRS Day-Night Band is a single broadband channel spanning roughly 500 to 900 nm, sensitive enough to detect moonlit clouds yet able to image city centres without saturating on most passes. That dynamic range comes from three on-chip gain stages that switch automatically depending on scene brightness. The result is a calibrated, physically meaningful radiance in units of nW cm⁻² sr⁻¹, which is what separates VIIRS from its predecessor DMSP-OLS. OLS had no radiometric calibration at all: its digital numbers cannot be converted to physical units, vary between satellites, and were deliberately saturated over bright cities to protect the detector.
The 750 m nadir resolution degrades toward the swath edge but the sensor's wide 3,000 km swath means every point on Earth is observed at least once per night, usually twice when both Suomi-NPP and NOAA-20 are counted. That daily cadence is the sensor's greatest analytical strength. A power outage, a new industrial facility, or a destroyed substation can appear in the record within 24 hours.
Saturation and the bright-city problem
The DNB's gain stages are designed to avoid saturation over most scenes, but dense urban cores in wealthy countries, particularly in East Asia, Western Europe and North America, still saturate the detector on clear nights. When a pixel saturates, you lose the ability to distinguish a moderately lit district from an intensely lit one. For economic-activity proxies or electrification mapping in already-bright cities, this matters: the sensor cannot tell you whether downtown Seoul got brighter between 2015 and 2023 if both epochs are saturated.
The practical workaround is to work at the urban fringe, to use monthly composites that filter out the brightest observations, or to switch to a different analytical question entirely, such as the spatial extent of lit area rather than its intensity. The Colorado School of Mines Earth Observation Group publishes monthly and annual composites with various stray-light and saturation masks applied, and these are the standard starting point for most urban-light studies.
Lunar contamination and the stray-light correction
The DNB is sensitive enough to detect moonlight reflected from clouds and snow. On nights near full moon, that background illumination can swamp faint anthropogenic signals, particularly in high-latitude regions where oblique solar and lunar geometry extends illumination into what should be the dark part of the orbit. NOAA and the Earth Observation Group publish stray-light corrected products that attempt to remove this contamination, but the correction is imperfect over sea ice and bright desert surfaces.
Analysts should always check the lunar phase and solar zenith angle metadata before interpreting a single-night DNB image. Monthly composites assembled by selecting the minimum-radiance observation per pixel across a month largely sidestep the problem, at the cost of temporal resolution. For conflict damage or disaster response, where you need a specific night's data, lunar phase is a genuine operational constraint.
Stitching DMSP to VIIRS without breaking the record
The DMSP-OLS archive runs from the late 1970s to roughly 2013 on multiple satellites, none of which were cross-calibrated with each other or with VIIRS. Researchers have published intercalibration methods, most notably work by Elvidge and colleagues at NOAA, that attempt to harmonise the two records by fitting regression relationships over stable, slowly changing reference areas. These methods reduce but do not eliminate the discontinuity. The transition period between the final DMSP satellites and Suomi-NPP is particularly problematic because the two sensors overlap only briefly and the OLS digital numbers in that period are degraded.
For any analysis that needs a consistent multi-decade trend, the honest position is to treat the pre-2012 DMSP record and the post-2012 VIIRS record as two separate datasets joined by an explicit intercalibration step, and to report the uncertainty that step introduces. Studies that simply concatenate the two without adjustment routinely produce artefacts that look like real economic or electrification trends but are sensor transitions.
Gas flares, fishing fleets and conflict damage: three analytic use cases
Gas flaring is among the most reliable DNB applications. Flares are bright, persistent, spatially fixed and occur in otherwise dark regions. The VIIRS Nightfire product uses shortwave infrared channels alongside the DNB to estimate flare temperature and radiant heat, which can be converted to flared gas volume estimates. The World Bank's Global Gas Flaring Reduction Partnership has used this data to publish country-level flaring inventories, making it one of the few satellite-derived datasets with direct policy uptake at the intergovernmental level.
Squid and saury fishing fleets in the western Pacific and Indian Ocean use powerful lights to attract their catch to the surface at night. These fleets appear as transient bright clusters in the DNB, often far from any declared fishing zone. Cross-referencing DNB detections with AIS vessel positions (covered separately in the AIS and dark-ship page) identifies vessels that are fishing without broadcasting their location. Conflict damage assessment is a third application: the DNB has been used to document electricity-grid collapse in Syria, Yemen and Ukraine by comparing pre-conflict and post-event composites. At 750 m resolution, individual neighbourhood-level outages are detectable, though the method cannot distinguish a destroyed substation from a deliberate blackout or a fuel shortage.
Satellize runs DNB-based electrification and economic-activity analytics on open VIIRS data as part of its satellite-data analytics service. Its crop-estimation programme for the Kingdom of Tonga, which combines optical and radar data, uses nighttime-light layers as a secondary indicator of rural infrastructure change.
What the sensor cannot do
Cloud cover blocks the DNB completely. In persistently cloudy regions, monthly composites may have fewer than five cloud-free observations in a given month, making trend detection unreliable. The 750 m pixel is too coarse to resolve individual buildings or streets, so sub-neighbourhood attribution is not possible. The sensor has no spectral discrimination within its single broad band, meaning it cannot distinguish sodium streetlights from LED arrays or natural gas flares from coal fires on spectral grounds alone. And because the DNB is a passive optical sensor, it provides no information on days when the sun is above the horizon, which sounds obvious but matters for any analyst trying to understand whether a facility is operating on day shifts only.
Typical figures
| Spatial resolution (nadir) | 750 m |
| Swath width | ~3,000 km |
| Revisit (single satellite) | Daily (one pass per night) |
| Revisit (Suomi-NPP + NOAA-20) | Two passes per night, ~50 min apart |
| Spectral band | 500–900 nm (panchromatic, single band) |
| Dynamic range | ~3×10⁻⁹ to ~2×10⁻² W cm⁻² sr⁻¹ (seven orders of magnitude across three gain stages) |
| Minimum detectable radiance | ~3×10⁻⁹ W cm⁻² sr⁻¹ (high-gain stage, clear sky) |
| Archive depth (VIIRS DNB) | October 2011 to present (Suomi-NPP); January 2018 to present (NOAA-20) |
| Heritage archive (DMSP-OLS) | Late 1970s to ~2013; ~2.7 km resolution, uncalibrated |
| Standard composite products | Monthly and annual cloud-free composites with stray-light and lunar correction, published by NOAA/Earth Observation Group |
Analytics Satellize can run
| Electrification change map | Pixel-level radiance differencing between baseline and current monthly composites, with cloud-fraction and lunar-phase quality filtering | GIS layer (GeoTIFF or vector polygons) showing newly lit or newly dark areas at 750 m resolution, with change magnitude in nW cm⁻² sr⁻¹ |
| Economic-activity proxy index | Sum of radiance (SoL) or lit-area extent aggregated to administrative boundaries, following published methods from the Elvidge/NOAA group and Henderson et al. (World Bank) | Time-series table by district or province, updated monthly, with explicit saturation-flag annotation for bright urban cores |
| Gas flare inventory and trend report | VIIRS Nightfire detection algorithm combining DNB and M-band SWIR; cluster persistence filtering to separate flares from transient fires | Georeferenced flare point dataset with estimated radiant heat and temporal activity profile; annual comparison report |
| Fishing-vessel light detection | Transient bright-cluster detection in nightly DNB imagery; spatial and temporal filtering to exclude persistent infrastructure; optional cross-reference with AIS positions | Nightly vessel-cluster alert feed with centroid coordinates, estimated radiance and cloud-cover confidence flag |
| Conflict infrastructure damage assessment | Pre/post composite differencing over conflict-affected areas; change significance tested against inter-annual variability baseline | PDF assessment report with annotated imagery and tabular summary of radiance change by settlement or grid zone |
| DMSP-to-VIIRS intercalibrated trend series | Regression-based intercalibration over stable reference zones following published Elvidge et al. methodology; uncertainty bounds propagated through the join | Long-record radiance time series (1992 to present where coverage allows) with explicit intercalibration uncertainty column |
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.