Nov 16, 2026
Field notes: the week the built-up mask found a subdivision the plan set didn't

We were three weeks into screening candidate alignments for a bypass segment, working off the usual stack: county parcel data, a topo sheet pulled from the state DOT archive, and a planimetric base that someone on the team swore was "basically current." Basically current turned out to mean four years old.
The southern alignment, the one with the gentlest grade and the fewest stream crossings, ran clean on paper. Open parcels, ag zoning, nothing flagged in the right-of-way corridor. It was the front-runner going into the next milestone review, the one we'd have put in front of the client as the recommended centerline pending field verification.
Then someone pulled a current built-up mask to cross-check impervious cover against the floodplain layer, mostly as a sanity check on drainage, and there it was: a cul-de-sac with eleven finished lots sitting almost dead center on the alignment. No subdivision on the plan set. No permit record we had access to. Just rooftops, driveways, and a stub street that didn't exist the last time anyone updated the base mapping.
This is the normal failure mode of base mapping that gets built once and trusted for years. Plan sets get refreshed on whatever cycle the agency's survey budget allows, which in a lot of counties is "whenever someone notices it's wrong." Meanwhile a developer closes on forty acres, grades it, and has slab on grade before the next aerial flight even gets scheduled. Nobody updates a parcel layer because a subdivision got built. Somebody updates it eventually, on their own schedule, usually after a complaint.
The part that stings is how avoidable the rework was. If that subdivision had surfaced during survey instead of during screening, it would have cost us field time, a redline, and an awkward conversation about why the recommended alignment needed to move after the client had already started picturing the southern route on a map. Catching it at the desk, before anyone mobilized a crew, is the entire point of scoring candidates the way we did. It just happened to be luck that someone ran the cross-check that week.
The fix isn't "check more carefully." Careful checking against stale data still produces a stale answer. The fix is cross-referencing against something with a shorter lag than the plan set, something pulled close to project start rather than inherited from whatever vintage the county's GIS shop last updated. A built-up mask derived from recent imagery doesn't care what's zoned or platted or recorded. It shows what's actually standing. That's a different kind of ground truth than a parcel layer, and it catches a different kind of mistake, the kind where the paperwork hasn't caught up to the construction crew.
We moved the alignment north a few hundred feet that week, which solved the rooftop problem and traded it for a steeper approach grade and one more wetland delineation to scope. That's a fair trade to make in week three of screening, with a felt-tip pen and three candidate lines on a desk. Making the same move in week nine, with a crew already mobilized and a landowner standing in their driveway asking why there are stakes in their side yard, costs a lot more than a steeper approach grade.
If you're scoring candidates the same way, run a current built-up check against every alignment before anyone picks a front-runner. That catches conflicts on alignments that look clean on paper, which is exactly where they hide. We run that check now as part of the same pass that scores land cover and slope together, built around a constraint-scored map of candidate alignments, so the subdivision shows up in week three instead of week nine.
Pull a current built-up mask before you circle a favorite alignment on the plan set.