What counts as built-up in a national land cover scheme

Every national land cover legend has a built-up class. Almost none of them define it the same way, and the gap between "built-up" and "impervious surface" is where most inter-agency arguments start.

Ask three analysts what goes in the built-up class and you'll get three defensible answers. One counts rooftops and pavement by spectral signature. A second counts parcels zoned residential or commercial regardless of canopy cover over them. A third counts anything with a structure footprint above some minimum size, dropping isolated farm sheds below threshold. All three are internally consistent. None of them agree with each other, which is a problem the moment your statistics office tries to report a national built-up area figure that planning and environment both cite without a footnote explaining whose rule it used.

Built-up vs impervious: not the same measurement

Impervious surface is a physical property: a pixel either sheds rainwater and reflects like concrete, asphalt, or roofing, or it doesn't. It's measurable from spectral response more or less directly, which is why a lot of legacy classification shortcuts treat "impervious" and "built-up" as interchangeable.

They're not. A gravel farm track is impervious but nobody calls it built-up. A park with 40% canopy over paved paths reads as vegetated in a coarse spectral classifier but sits inside a built-up parcel in any cadastral sense. A solar farm is almost entirely impervious at ground level and spectrally flat, but most legends file it under a generation or industrial subclass rather than residential/commercial built-up, because that's what the planning department actually wants to track separately.

The schemes that hold up over multiple annual editions treat built-up as a land use judgment informed by impervious cover, not a direct readout of it. Impervious fraction is one input. Structure footprint, parcel context, and a minimum mapping unit do the rest of the work.

Where the legend has to make a call

A handful of recurring cases are where classification schemes either get specific or get argued about every cycle:

• Transport corridors. A four-lane highway is impervious for kilometers but most national schemes carve it into its own linear class rather than folding it into built-up, because lumping it in inflates built-up area in rural districts that have one road and nothing else.
• Quarries and extraction sites. Bare, graded, and often fenced like an industrial site, but not built-up in any land use sense. Needs its own bare/extractive subclass or it pollutes the built-up trend line every time a new quarry opens.
• Informal and low-density settlement. Structure footprint can sit well under the minimum mapping unit that works fine for a planned subdivision, so a fixed-threshold rule systematically undercounts settlements a census would count as built-up. This is usually the single biggest source of disagreement between a land cover team and a population/housing statistics team working from the same year's imagery.
• Greenhouses and covered agriculture. Spectrally near-impervious, structurally a roof, but agricultural by use. Most legends put these in cropland or a dedicated glasshouse subclass, explicitly excluded from built-up, which only works if the rule is written down and applied the same way in the north region as the south.

None of these are hard to resolve individually. What's hard is resolving them the same way at the next annual update, when a different analyst or a different contractor is running the classification and nobody has the previous year's edge-case log in front of them.

Why the rule has to survive the handoff

A built-up definition that lives in one analyst's head, or in a methodology note from a contract that ended two years ago, doesn't survive a staff change or a new vendor. The next edition either reinvents the threshold rules from scratch or silently drifts from the last one, and a drift nobody flagged looks exactly like a trend nobody can explain when the statistics office asks why built-up area jumped nine percent in a district that didn't build anything.

That's the argument for running the classification and the delta-check as one recurring process rather than one more ad-hoc contract. Land Use Mapping applies a consistent built-up definition against each year's imagery and flags what moved class against the prior edition, so the number your statistics office cites is the same one planning is looking at.

If your agency is still reconciling two or three versions of "built-up" across departments, it might be worth seeing what one shared layer looks like.