Port and freight terminal activity indexed by nighttime light intensity
VIIRS Day/Night Band radiance turns port floodlighting into a continuous operational index. The method is cheap, globally consistent, and honest about its 750 m resolution floor.
Sensors
- VIIRS Day/Night Band (DNB): Suomi-NPP and NOAA-20 each provide daily global coverage at approximately 750 m ground sample distance. DNB detects radiance from roughly 3×10⁻⁹ to 2×10⁻² W cm⁻² sr⁻¹, sensitive enough to resolve individual floodlit berths at major ports, though adjacent facilities within the 750 m footprint blur together.
- DMSP-OLS (historical archive): Operational from 1992 to 2013 across multiple satellites. Coarser at ~2.7 km and prone to saturation over bright urban cores, but provides three decades of baseline against which VIIRS trends can be anchored. Useful for constructing long-run port-activity histories.
- VIIRS Nightfire (VIIRS VNF product): A separate VIIRS product from Colorado School of Mines that isolates high-temperature combustion sources using shortwave infrared bands. Helps separate gas flares and vessel exhausts from floodlight signals in busy anchorage areas, reducing contamination of the port radiance index.
- MODIS Terra/Aqua (contextual): 500 m daytime imagery used to delineate port facility footprints and mask water pixels before applying the nighttime radiance model. Not a light-detection sensor, but essential for defining the spatial mask that constrains which pixels enter the activity index.
Why a floodlight is an economic signal
Port operations run on electricity. Gantry cranes, reach stackers, reefer plugs, and perimeter lighting all draw power whenever cargo moves. That power radiates upward. The VIIRS Day/Night Band, flying aboard Suomi-NPP at roughly 824 km altitude, reads that upwelling radiance every night at a noise-equivalent delta radiance of around 2×10⁻¹⁰ W cm⁻² sr⁻¹, which is sensitive enough to detect a single football stadium's worth of floodlighting from orbit.
Published economic research, including work using Chinese port data and global shipping indices, has documented statistically significant correlations between DNB radiance and container throughput, vessel call frequency, and GDP proxies for coastal economies. The intuition is straightforward: a terminal running three shifts burns more light than one running one. The satellite does not care whether the port authority publishes its throughput figures or not.
What 750 metres actually means for port analysis
The DNB's ground sample distance of approximately 750 m is the method's most important honest constraint. A pixel that size covers roughly 56 hectares. Many mid-sized ports fit inside two or three pixels. That means the radiance index measures a facility cluster, not an individual berth or terminal operator. Where a container terminal, a bulk terminal, and a fishing harbour sit within a kilometre of each other, their signals mix irretrievably in the DNB.
The practical workaround is to define port zones using high-resolution daytime imagery, assign fractional pixel weights based on land-cover type, and then treat the resulting radiance as a zone-level index rather than a point measurement. This works well for large, spatially isolated ports. It works poorly for dense, multi-operator waterfront districts where the spatial mask is ambiguous. Buyers should be clear about which ports they need indexed before assuming the method applies uniformly.
Fishing fleets, moonlight and other sources of noise
Squid fishing fleets use high-intensity lights to attract their catch. In the East China Sea, the Sea of Japan, and the waters off South America's Pacific coast, these fleets produce radiance plumes that can rival or exceed coastal port signals, particularly between June and October. When a fishing ground sits near a port anchorage, the fleet's sky-glow inflates the apparent port-area radiance, making a quiet port look busier than it is.
Moonlight is a subtler problem. The DNB is calibrated to work across the full lunar cycle, but scattered moonlight off cloud decks raises the background radiance floor and compresses the dynamic range available for detecting facility-level signals. Standard practice is to flag high-lunar-illuminance nights and either exclude them from the time series or apply a lunar correction model. Cloud cover is a harder problem: an overcast night produces no usable DNB observation at all. At mid-latitudes, cloud-free acquisition rates can fall below 50 percent in winter months, creating gaps that require interpolation or compositing before a smooth activity index can be constructed.
Disaster recovery arcs: what shutdown and restart look like from orbit
Port shutdowns following natural disasters produce sharp, dateable drops in DNB radiance that are visible in the raw time series without any modelling. After Typhoon Hainan (2013) struck the Philippines, nighttime light data showed radiance collapse across affected coastal areas within 24 hours of landfall, with partial recovery curves that tracked infrastructure restoration over subsequent weeks. Similar patterns have been documented after the 2010 Haiti earthquake and the 2011 Tōhoku tsunami, using both DMSP-OLS and VIIRS depending on the event date.
For logistics planners and insurers, this recovery arc is the operationally relevant signal. The question is not just whether a port is dark, but how quickly radiance returns to pre-event baseline and whether it plateaus below that baseline, which would suggest structural capacity loss rather than temporary closure. A 30-night rolling composite, updated nightly, can track that arc in near-real-time. The International Charter on Space and Major Disasters has used nighttime light data as a rapid damage proxy in several activations, which gives the method a degree of operational validation.
Building a defensible activity index
A raw DNB radiance time series is not an activity index. Constructing one requires several steps. First, define the spatial mask for each port zone using daytime optical imagery. Second, compute the mean or sum of cloud-free, low-lunar-illuminance DNB pixels within each mask for each night. Third, apply a seasonal decomposition to remove the fishing-fleet and agricultural-burn signals that recur on annual cycles. Fourth, normalise against a reference period to produce a dimensionless index that is comparable across ports of different absolute brightness.
The resulting index typically has a temporal resolution of seven to fifteen days in practice, once cloud gaps are filled by compositing. That is not daily. It is, however, global, free at the point of acquisition from NOAA's archive, and available with a latency of roughly 24 hours for the raw DNB product. For ports that publish no throughput data and host no AIS infrastructure, it is often the only continuous economic signal available at all.
Satellize runs this class of analysis as part of its broader satellite-data analytics practice. The workflow is the same family of open-sensor methods used in the Tonga crop-estimation programme: define zones, ingest open archive data, apply validated signal-processing, and deliver a structured output rather than raw imagery.
Where the method earns its keep and where it does not
The strongest use cases are monitoring ports in jurisdictions that suppress or delay throughput statistics, tracking recovery after disruption when ground access is restricted, and building a long-run baseline for economic analysis in coastal emerging markets. For these applications, the 750 m resolution is a manageable constraint and the global, free archive is a decisive advantage.
The weakest use cases are distinguishing between two adjacent terminals operated by competing shipping lines, detecting short-duration events like a single vessel call, or measuring activity at small inland freight depots whose footprint falls within a single DNB pixel alongside unrelated urban lighting. For those questions, synthetic aperture radar or high-resolution optical imagery, covered in the sibling pages on container yard occupancy and vessel dwell time, will give more reliable answers. The honest position is that nighttime lights work best as a macro index and poorly as a micro-measurement tool.
Typical figures
| Spatial resolution (DNB) | ~750 m ground sample distance; pixel area ~56 ha |
| Revisit frequency | Daily (Suomi-NPP + NOAA-20 together provide two passes per night at most latitudes) |
| Effective temporal resolution of index | 7–15 days after cloud compositing at mid-latitudes; shorter in tropical dry seasons |
| Raw data latency | ~24 hours for VIIRS DNB daily product from NOAA archive |
| Spectral band | DNB: panchromatic 500–900 nm; radiance range ~3×10⁻⁹ to 2×10⁻² W cm⁻² sr⁻¹ |
| Historical archive depth | VIIRS from 2012; DMSP-OLS from 1992 (coarser, saturates over bright ports) |
| Minimum detectable signal | Noise-equivalent delta radiance ~2×10⁻¹⁰ W cm⁻² sr⁻¹ under favourable conditions |
| Coverage | Global, every night; polar regions receive multiple passes per night |
| Key confounders | Cloud cover, moonlight (>30% illumination), fishing fleet sky-glow, gas flares |
| Delivery formats | GeoTIFF composites, CSV time-series per port zone, GIS polygon layers |
Analytics Satellize can run
| Port activity index time series | Cloud-masked DNB radiance compositing with seasonal decomposition, following methods published in the remote sensing economic literature | Monthly CSV and chart pack per port zone, updated weekly |
| Disruption alert | Threshold detection on 7-night rolling radiance composite against 90-day baseline; flags drops exceeding two standard deviations | Automated alert with radiance anomaly map as GeoTIFF |
| Recovery arc report | Post-event radiance trajectory fitted against pre-event baseline to estimate percentage capacity restoration over time | PDF report with time-series chart and estimated recovery date range |
| Multi-port comparative ranking | Normalised radiance index across a defined portfolio of ports, enabling relative activity comparison independent of absolute facility brightness | Interactive dashboard or static GIS layer updated on agreed cadence |
| Fishing-fleet contamination flag | Cross-reference with VIIRS Nightfire vessel-detection product to identify nights where anchorage fishing activity inflates port-zone radiance beyond a defined threshold | Quality-flag column appended to activity index CSV |
| Long-run baseline construction | Inter-calibrated DMSP-OLS to VIIRS DNB time series using published cross-sensor normalisation coefficients, extending the record back to 1992 | Annual index values per port zone from 1992 to present as CSV |
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.