Attributing a night-light increase to your project, not the one next door

Aerial dusk view of a small town where a lit road stretches out from the built-up area into the surrounding dark countryside.

You're looking at the quarterly brightness extract for your corridor and there's a bloom right along the edge of your draw area. Good news, except the national utility also has a feeder extension going into the next district, and you can't immediately tell whose rollout lit up those pixels. This is the attribution problem that comes up on almost every grid-extension loan once you start using satellite light as a between-survey check: brightness doesn't carry a project ID.

Why edge pixels are the problem, not the whole scene

At coarse resolution, a single pixel can sit on top of more than one administrative boundary, and light bleeds across that boundary the way sound carries across a fence line. A settlement 2-3 km from your nearest energized transformer can show a brightness bump that's actually glow from a neighbouring programme's substation, not anything your disbursement paid for. The middle of your corridor, where your network is dense and nobody else has infrastructure, is rarely the problem. It's the border settlements and the contested service areas, places two programmes both claim on paper, where the signal gets ambiguous.

This matters because a false positive here isn't academic. If a brightness increase at the edge of your area gets logged as project progress and it turns out to be spillover from the adjacent programme's rollout, your next field mission and your portfolio note are both built on the wrong number.

Separating your rollout from theirs

A few checks catch most of the ambiguous cases before they reach a report:

  • Timing against your own energization dates. Pull the commissioning date for the nearest substation or feeder segment in your project files. If the brightness increase in a cell starts months before that date, it isn't your rollout lighting it up.
  • Shape of the bloom. Light spreading out from your known network nodes tends to grow outward from those points over successive months. A neighbouring programme's rollout usually has its own separate center, often with a different grid orientation, and it will keep growing from its own node, not yours, even as the two blooms edge closer together.
  • Cross-reference with the other programme's schedule, if you have it. Development programmes financed by other institutions often publish their own completion milestones. A quick check of their rollout timeline against your brightness bump settles a surprising number of these cases on its own.
  • Treat genuinely ambiguous cells as flagged, not counted. If a settlement sits within the plausible spillover radius of both networks and the timing doesn't clearly favor one, don't fold it into your progress number either way. Flag it for the field team and move on.

What this resolution can and can't tell you

Worth saying plainly: at ≥250 m, panchromatic, monthly cadence, you are not resolving individual households or even individual streets. You're reading a settlement-level glow, which is exactly the right instrument for catching "something changed here since last quarter" and exactly the wrong instrument for a courtroom-grade attribution claim. The honest use of this data is as a trigger, not a verdict. It tells you where to send the verification visit or where to ask the implementing agency a pointed question before the formal survey comes due. It doesn't replace that survey, and treating a monthly light read as proof of attribution is asking more of a 250 m pixel than it can give.

A programme-area boundary file, loaded against the monthly read, does most of the heavy lifting here. If you're drawing hard conclusions from a single bright cell at the edge of your corridor without that boundary and timing check, you're more likely attributing someone else's feeder than your own.

Electrification Map runs exactly this kind of monthly night-light read over your programme area, so edge-of-corridor ambiguity is something you catch between surveys instead of something you discover at completion review.

Start a project

Back to the blog