Field Boundary Map

Reading a shapefile like tree rings

Old wooden fence post tangled with dried vines at the edge of a harvested field under heavy fog

A colleague sent over a parcel layer last month with a note that just said "see if you can tell what happened here." I opened it in QGIS, zoomed to a cluster of fields near the river, and started laughing, because I could. Three editors, maybe more, each one visible in the geometry like growth rings in a stump.

The oldest polygons were the loose ones. Big, confident vertices spaced twenty or thirty meters apart, boundaries that cut straight through what should have been a curved drainage line. That's late-1990s digitising off low-res scans, back when a mouse click cost you real effort and nobody was zoomed past 1:10,000. You trust those boundaries the way you trust a hand-drawn property sketch from a deed book: roughly right, wrong in the details that matter now.

Then a layer of parcels from what I'd guess was the mid-2000s, all suspiciously square. Someone had a habit of snapping to orthogonal angles even when the actual field edge curved with the contour. You can tell because the real crop line in the background imagery drifts away from the polygon edge by a few meters on one side and overlaps it on the other. That's a digitiser who learned on CAD before they learned on aerial photos, and the instinct never left.

And then the recent stuff. Dense, twitchy, a vertex every meter or two, following every little notch where a tractor turned wide around a drainage ditch. Technically more accurate. Also heavier than the database needs, and in a couple of spots you can see where the hand holding the mouse got tired, because the vertex spacing gets sloppy toward the end of a long boundary run.

None of this is unusual. If you've maintained a parcel registry for more than a decade you've probably got the same thing sitting in your gdb right now, you just haven't gone looking. Every boundary layer is partly a record of the land and partly a record of who sat at the workstation that year, what software they had, how they'd been trained, and how many parcels they still had to finish before the Friday deadline. The land doesn't change its editing style. The staff do.

This matters more than it sounds like it should, because area calculations inherit all of it. A field digitised with loose, generalized boundaries in 1998 and never touched since can carry an area figure that's off by a meaningful margin against what a combine covers today, especially where field edges have crept or a ditch got rerouted. When that number feeds a subsidy payment or sits behind a loan collateral valuation, "roughly right" stops being good enough, and nobody wants to be the one explaining to an auditor why three adjoining parcels have three different vertex densities and two different coordinate drifts.

Chasing down which editor did what and re-tracing their work by hand isn't realistic. That's a career, not a task. Treating each annual cycle as a clean pull instead of a patch on top of the old lines is what works: pull current imagery, generate the boundary fresh, and let the new layer stand on its own instead of inheriting someone else's 2006 habits. That's the whole premise behind the boundary layer Field Boundary Map builds from each year's imagery: it doesn't care what style the last person used, because it isn't editing their lines, it's drawing new ones straight from the pixels.

Somewhere down the line, whoever opens this year's shapefile in 2040 is going to notice a shift in the data too. Might as well make sure what they find is one consistent method, not another decade of handwriting to decode. If your registry is overdue for that kind of reset, that's what we're here for.

Get started

← Back to the blog