Night-light intensity monitoring of remote mining camp activity
VIIRS and commercial low-light sensors detect radiance from remote mining camps nightly, giving operational-status signals where cloud and access deny conventional observation. Radiance is a proxy, not a headcount, and flare contamination is a real problem.
Sensors
- VIIRS Day/Night Band (DNB): 750 m nadir resolution, nightly global coverage from Suomi-NPP and NOAA-20/21. Detects radiance from roughly 3×10⁻⁹ W cm⁻² sr⁻¹ upward. Free archive from 2012. Coarse enough that a small camp can be confused with a nearby flare or road corridor.
- Landsat 8/9 OLI Panchromatic: 15 m resolution, but a 16-day revisit and no dedicated low-light mode. Occasional nightside acquisitions over high-latitude sites in winter have been used in published studies to map lit infrastructure at much finer spatial detail than VIIRS, though coverage is opportunistic.
- Planet SkySat: 0.5 m panchromatic, tasked on demand. Not a dedicated low-light sensor, but short-exposure nightside imagery has been demonstrated for bright industrial targets. Useful for confirming spatial extent of lit areas identified by VIIRS, subject to tasking cost and cloud.
- Lumos low-light constellation: Commercial constellation designed specifically for nocturnal radiance mapping at sub-100 m resolution. Revisit and archive depth are still maturing as of 2024; treat published specifications as indicative rather than operationally proven at scale.
What a lit camp actually tells you
A functioning mining camp consumes electricity around the clock: accommodation blocks, mess halls, processing plant lighting, security perimeters, and equipment charging. That consumption produces upward radiance detectable from orbit. The inferential chain is: radiance observed, power consumption inferred, operational status inferred from that. Each step introduces uncertainty.
VIIRS DNB measures top-of-atmosphere radiance in a single broadband visible-to-near-infrared channel (roughly 500–900 nm). It does not measure watts consumed on the ground. Atmospheric scattering, lunar phase, snow cover and thin cloud all modulate the signal. A camp that appears brighter in January than in July may simply be reflecting more moonlight, not running more shifts. Analysts must account for lunar illumination and flag cloud-contaminated pixels before drawing any operational conclusion.
Flares are not camps, and camps are not headcounts
Processing plants at gold, copper and oil-associated mining operations frequently burn off waste gas. A flare stack can saturate VIIRS DNB pixels and bleed radiance into adjacent pixels that nominally cover accommodation areas. Published work on Nigerian and Permian Basin flaring shows that a single large flare can produce radiance an order of magnitude above a fully lit camp of several hundred people. Separating the two requires either higher-resolution imagery to distinguish point sources spatially, or temporal analysis: flares tend to be persistent and spectrally hot, while camp lighting follows shift patterns.
Even when flare contamination is ruled out, radiance does not convert cleanly to workforce numbers. A skeleton crew maintaining equipment overnight may leave all lights on. A full production workforce in a well-managed camp may use motion-sensor lighting. The honest position is that radiance tracks operational status, meaning active versus dormant, and relative changes in activity level, not absolute headcount. Clients who need headcount require a separate data source, such as mobile-device signal density or helicopter manifest records.
The cloud problem and how nightly revisit partly solves it
Many high-value mining regions sit under persistent cloud: the Congo Basin, Indonesian archipelago, parts of the Andes. A single VIIRS overpass on a cloudy night yields nothing usable. This is the central operational constraint. The mitigation is temporal aggregation: monthly composites that retain only clear-sky observations can reconstruct a reliable radiance baseline even where individual nights are mostly obscured. The Colorado School of Mines VIIRS Nightfire programme and NOAA's VIIRS monthly composites, both publicly archived, demonstrate this approach.
Monthly composites smooth out shift-pattern variation, though. A mine that ran at full capacity for three weeks and shut down in the final week of the month will look healthier than it is. For clients who need week-level resolution, the answer is to use raw nightly DNB data and accept a higher rate of cloud-contaminated gaps, flagging them explicitly rather than interpolating through them.
Combining sensors to sharpen the inference
VIIRS provides the temporal backbone: nightly, global, free. Commercial sensors fill the spatial gap when the question changes from 'is this camp active?' to 'which specific structures are lit and what is the spatial footprint of activity?'. A standard workflow starts with VIIRS to flag anomalous radiance change, then tasks a high-resolution optical or SAR sensor to characterise the cause. SAR is worth mentioning here even though it is not a low-light sensor: Sentinel-1 C-band imagery at 10 m resolution can detect vehicle and equipment density in camp yards regardless of cloud or darkness, and pairing it with DNB radiance reduces ambiguity considerably.
Landsat nightside acquisitions, where available at high latitudes in winter, offer an intermediate option at 15 m and no tasking cost. The USGS archive contains a modest but growing set of these scenes. They are not systematic enough to build a revisit cadence around, but they serve well as confirmatory snapshots when a VIIRS anomaly needs spatial attribution.
Honest limits before you commit to a monitoring programme
VIIRS at 750 m will not resolve individual buildings. A camp smaller than roughly one to two VIIRS pixels (under about 500 m across) may be undetectable against background noise unless it is unusually bright. The published minimum detectable radiance for DNB under good conditions is around 3×10⁻⁹ W cm⁻² sr⁻¹, but practical detection thresholds in cluttered or moonlit scenes are higher. Small artisanal camps are frequently invisible; that monitoring gap belongs to a different page in this library.
Commercial low-light sensors such as Lumos promise sub-100 m resolution, which would resolve individual structures. As of 2024, the constellation's revisit consistency and archive depth are not yet publicly documented to the same standard as VIIRS or Sentinel. Programmes that depend on this data should budget for a validation phase rather than assuming specification-sheet performance.
Satellize runs VIIRS DNB time-series analytics on open archive data and can add commercial tasking for spatial confirmation under client licence. The methodology is the same class used in the Tonga crop-estimation programme: change detection against a calibrated baseline, with explicit uncertainty flags rather than false precision.
What a monitoring programme actually delivers
The practical output for an investor, regulator or operator is a weekly or monthly operational-status signal for each monitored site: active, reduced, dormant, or data-gap. Trend lines over six to twelve months reveal whether a camp is ramping up, holding steady, or winding down. Sudden drops in radiance, especially when correlated with known weather events or commodity price movements, become a structured data point rather than an anecdote from a field contact.
The signal is most credible when it confirms or contradicts other data streams: shipping manifests, power purchase records, or publicly filed production reports. Used in isolation, it is suggestive. Used as one layer in a multi-source picture, it is genuinely useful for decisions about capital allocation, royalty auditing, or supply-chain due diligence.
Typical figures
| Primary sensor spatial resolution | VIIRS DNB: 750 m nadir. Lumos: sub-100 m (indicative). Landsat nightside: 15 m panchromatic (opportunistic). |
| Revisit frequency | VIIRS: nightly (two satellites, Suomi-NPP and NOAA-20/21). Commercial low-light: varies by constellation and tasking. Landsat nightside: non-systematic. |
| Spectral channel | VIIRS DNB: broadband ~500–900 nm. Landsat OLI Pan: 500–680 nm. Low-light commercial: typically panchromatic visible. |
| Minimum detectable radiance (VIIRS DNB) | ~3×10⁻⁹ W cm⁻² sr⁻¹ under ideal conditions; practical threshold higher in cluttered or moonlit scenes. |
| Cloud penetration | None for optical/DNB. SAR (Sentinel-1) used as complementary all-weather layer for structural detection. |
| Archive depth | VIIRS: from 2012 (Suomi-NPP). Landsat 8 nightside: from 2013 (sparse). Commercial low-light: limited pre-2022 archive. |
| Latency (VIIRS public archive) | Typically 1–3 days for standard products via NASA FIRMS and NOAA STAR portals. |
| Typical analysis cadence | Weekly or monthly composite; nightly raw data available but cloud-gap rate may exceed 50% in tropical regions. |
| Delivery formats | GeoTIFF radiance rasters, CSV time-series per site, GIS polygon overlays, PDF status reports. |
Analytics Satellize can run
| Operational status classification per site | Threshold and change-point detection on VIIRS DNB monthly composites against a 12-month rolling baseline | Monthly status report (active / reduced / dormant / data-gap) per monitored camp, delivered as PDF and structured CSV |
| Radiance trend index | Linear regression and anomaly scoring on cloud-cleared nightly DNB time series; lunar correction applied using published NOAA monthly compositing methodology | Site-level trend chart and index score, updated monthly, as GIS layer and tabular feed |
| Flare contamination flag | Spectral temperature screening using VIIRS M-band thermal channels (M10, M11, M13) to distinguish combustion sources from reflected artificial light, following Colorado School of Mines Nightfire methodology | Per-pixel flare flag appended to radiance time series; contaminated observations excluded from camp-activity index |
| Spatial footprint mapping of lit infrastructure | High-resolution optical or commercial low-light imagery segmentation to delineate lit building clusters, roads and plant areas within VIIRS-flagged zones | GeoJSON polygon layer of lit-area extent, updated on tasking trigger from VIIRS anomaly alert |
| Multi-source activity composite | Fusion of DNB radiance index with Sentinel-1 SAR backscatter change (vehicle and equipment density proxy) to reduce single-sensor ambiguity | Weekly composite score per site combining optical and SAR signals, with confidence band, as GIS layer |
| Anomaly alert | Automated threshold trigger when radiance departs more than two standard deviations from the rolling baseline on cloud-clear nights | Email or API alert within 48 hours of anomalous overpass, with supporting imagery thumbnail and cloud-cover metadata |
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.