From Signal to Standard Work, Part 1 of 4. One upstream operator, one year of multi-sensor emissions data.
This is the first of four parts. Across the series we follow one upstream operator through a single year of multi-sensor emissions data, from the messy first months to the point where the work finally paid off. If you run a detection program, the first year here will probably look familiar. The argument of the series is that it shouldn’t have to be anyone’s first year twice.
One upstream operator’s emissions program processed 1,039 detections last year, across five different sources: aerial surveys plus four satellite providers. You’d think the hard part would be analyzing all of that. As we’ve argued before, measuring was never the bottleneck. The hard part turned out to be agreeing on what was real.
Most of the false alarms are familiar
About one in five detections was a false alarm. What made that number hard to act on was the mix underneath it.
Half were the familiar false positives, where a crew rolls out and finds nothing. The other half were real emissions that simply weren’t the operator’s to count or act on. The gas belonged to a working-interest partner, or to a site the operator had sold months earlier that never left the registry. A few were pinned to the wrong location entirely.
The finding
Half the false alarms were real methane, just not the operator’s to act on. It came from divested sites and partners’ acreage the asset records had never caught up with.
The sources aren’t equally reliable
Satellite detections came back with a 52% false-alarm rate, while aerial ran at 17%. That threefold gap means a satellite ping and an aerial hit are not the same kind of evidence, and somebody has to decide how much weight each one carries before any of it adds up to a decision.
Separately, among the emissions that were confirmed real, about one in five couldn’t be categorized at all. That is a completely different measure from the false alarm rate because the detection was genuine, but the field notes did not carry enough detail to describe what had happened on site.
Every operator goes through this
None of this reflects badly on the operator. Asset registries drift, acreage trades hands, and operators come and go on sites faster than any map keeps up with. Every operator that brings on multi-sensor detection spends most of the first year on this same work: reconciling the equipment inventory, remapping asset references, pruning divested sites, teaching the system whose gas is whose.
The data is all there. What it lacks is context, and today operators close that gap by hand. People run SQL against data exports, asset spreadsheets get passed around on shared drives and inboxes, and every quarterly review circles back to the same question nobody quite answers, what is the plan moving forward?
The work is real and has to happen. Whether it happens by hand is a different question.
What comes after
Once reality is reconciled, you get a picture of where the gas actually is. It surprised the people who commissioned it, because while it confirmed some expectations, it flipped others on their head. That picture, and its limits, is Part 2.
And if the first year described here sounds like the year you’re in, the short version of this series’ argument is that it doesn’t have to be anyone’s first year anymore. Part 4 gets there.