Facility Expansion

2026-11-02

Field notes: three years of watching one gate, and the week it opened

A wooden perimeter fence with a gate at the crest of a grassy hill, beside a paved road curving away, under a stormy sky.

For three years the gate on the northeast perimeter did nothing. Same hardstand, same two bollards, same faint tire track curving off toward the service road and stopping short of the fence line like it had thought better of it. I pulled the tile every Monday because that was the rotation, not because I expected anything. You get used to a site that doesn't move. You start reading the shadows more than the structures, just to have something to log.

That gate is the kind of feature that trains you wrong. Three years of nothing teaches you that nothing is the baseline, and baselines are dangerous precisely because they're comfortable. I'd catalogued every innocuous explanation for a closed gate at a hardened site: seasonal access restrictions, a maintenance contract that never got renewed, a perimeter road rerouted two kilometers north for reasons that had nothing to do with intent. None of it was wrong. All of it was also irrelevant the week the gate opened, because by then the tire track had become a graded road, there was a laydown area where the scrub used to be, and a stack of what looked like precast sections sitting on pallets inside the fence.

The thing nobody tells you about long-dwell monitoring is that the change, when it comes, doesn't announce itself. It arrives as a pixel count that's slightly off from last week's, on a tile that looks ninety percent identical to fifty other tiles you've closed out without a second look. I almost didn't flag it. The gate itself hadn't moved. What had moved was everything twenty meters behind it, and I only caught that because I happened to pull the prior week's archive side by side instead of trusting memory, which after three years of a quiet site is exactly the habit that erodes.

That's the part worth writing down, because it's not really about this gate. It's about what three years of a static feature does to the person watching it. You stop treating the weekly pull as an inspection and start treating it as a formality. The site trains you to look past it. And the regime you're supposed to be enforcing doesn't care how many quiet Mondays came before the one that wasn't.

I've started running the comparison a different way since then. Less "did the gate change" and more "did the ground state around the gate change," because ground state is where the early signal lives, weeks before anyone pours a slab. Grading. Spoil. A new vehicle pattern on soil that used to be undisturbed, none of which requires the gate to open. It just requires someone to be looking at the right layer of the imagery and not get lulled by the feature everyone agreed to watch.

Three years of manual side-by-sides is also a long time to trust your own eye on something that's supposed to be defensible. The week the gate opened, the thing that actually closed the file wasn't my recollection of what the site used to look like. It was being able to put last month's tile and this week's tile in front of someone else and point at the excavation footprint without arguing about whether the shadows were doing something tricky. That's the gap a standing weekly watch on a named site is built to close, and it's why I don't run this kind of comparison from memory anymore, not on a site I care about.

Worth logging, for whoever inherits this site next: the quiet years are not the baseline. They're just the years before the week you were watching for.

Get started

← Back to the blog