What counts as a flare versus a well pad light in night-time imagery
Anyone who has pulled a VIIRS or Nightfire scene over a basin knows the first problem isn't finding bright pixels. It's figuring out which bright pixels are combustion at all. A flare stack, a well pad's security lighting, a drilling rig under full floodlights, and a compressor station's yard lights all show up as the same thing to an untrained eye: a dot of light on a dark background. For a regulator cross-checking operator-reported flare volumes, or a fund analyst using flare counts as a production proxy, getting that classification wrong means the whole downstream number is wrong too.
Why the distinction matters for emissions reporting
Flare counts get used as a stand-in for things operators don't always report promptly: how much associated gas is being burned off, whether a field is still in active development, whether a company's stated flaring-reduction targets match what's happening at night. If a pad light or a rig's work lights get logged as a flare, the count inflates. If a small, intermittent flare gets dismissed as background noise, the count understates actual burning. Either error moves an ESG flaring-intensity metric or a state regulator's compliance tally in the wrong direction, and neither is a rounding error when it feeds into a filing.
The physical tells that separate a flare from a pad light
A flare is combustion. That matters because combustion has a thermal signature that ordinary lighting doesn't. A few things to check before calling a light source a flare:
- Spectral behavior. Flares burn hot enough to show a strong short-wave infrared return, not just a visible-light glow. Pad lighting, yard lights, and rig floods are typically sodium or LED fixtures. They light up the visible band but don't carry the same heat signature in SWIR.
- Persistence and pattern. A pad's security lights come on at dusk and go off at dawn, every night, at a fixed location, with steady intensity. A flare's output tracks production: it can flicker, pulse, or vary in apparent size night to night as gas volumes change, and it can go dark entirely when a well is shut in or gas is routed to sales instead of being burned.
- Position relative to the pad layout. Flare stacks sit at a known offset from the wellhead, usually at the edge of the pad, away from tanks and the access road. A security light or yard light sits closer to equipment or along the perimeter fence line. Over a few nights of repeat passes, that offset is usually consistent enough to separate the two.
- Co-location with other infrastructure. A compressor station or a gas plant will show multiple light sources clustered tightly together, often including true combustion sources like a plant flare alongside ordinary facility lighting. Isolating the combustion pixel from the rest of the cluster takes a closer look at intensity and spectral signature, not just counting bright spots.
Where false positives creep in
A drilling rig is the most common source of false positives, not a well pad's security lighting. A rig under active operation runs floodlights at an intensity that can rival a small flare in a coarse-resolution scene, and rigs move between pads on a schedule that doesn't match any production pattern. A second common source is flaring at a facility that's been misclassified as a well site when it's actually a tank battery or gathering point, which changes what the light represents even if the combustion call is correct. A third is urban or oilfield-camp lighting bleeding into a monitoring area at its edge, especially along access roads with sodium lighting that reads bright in low-resolution passes.
None of this is reason to throw out night-time imagery as a flaring proxy. It's reason to run the same monitoring area night after night and build a baseline of what's a fixed light and what moves or varies with production, rather than calling a one-night bright spot a confirmed flare. A log built from repeated passes, checked against known infrastructure, catches a lot of what a single still image can't. That's the gap between a snapshot and a working flare count, and it's the reason a nightly flare log beats relying on one clear night and a guess.
If your current flaring numbers come from a single image or an operator's own filing, it may be worth comparing them against a few weeks of nightly reads before the next reporting cycle.