CORINE Land Cover vs a national bespoke classification: where the two legends pull apart
If your agency reports to Copernicus under CORINE and also publishes a national land cover product, you already know the two don't map onto each other cleanly. One is built for pan-European comparability. The other is built for whatever your statistics office, forestry service, or planning department actually needs counted. Here's where they diverge and what that costs you in practice.
What CORINE's nomenclature gives you
CORINE Land Cover (CLC) runs a three-level hierarchy topping out at 44 classes at Level 3: arable land, permanent crops, pastures, forest types split only to broadleaved/coniferous/mixed, and so on. The minimum mapping unit is 25 hectares for areal features and 100 m width for linear ones, refreshed on a five-year cycle (2000, 2006, 2012, 2018, with 2024 following the same schedule). That coarseness is deliberate. It's what lets the EEA stack your return against every other member state's without a national legend getting in the way.
For environmental reporting obligations, that's exactly what you want: a class scheme nobody has to translate. For anything your own ministries ask for, it's usually not enough. A 25-hectare MMU can't tell your agriculture department which 2 hectares are under vineyard versus orchard. A broadleaved/coniferous split can't tell your forestry service which stands are due for thinning.
Why agencies build a second, national legend anyway
So most national mapping agencies run a parallel bespoke scheme alongside CORINE: a finer MMU (often 0.5 to 1 hectare), crop-type and forest-type breakdowns that match whatever your agricultural census or forest inventory already uses, and built-up classes split by density because your urban planners care about that distinction even if CORINE doesn't. That national legend answers the questions your domestic users ask. CORINE answers the one Brussels asks.
The trouble starts when both products come from separate contracts, separate imagery runs, and separate delivery calendars. Your statistics office publishes cropland figures from the national product in March. Your environment unit submits a CORINE-compatible return in June, built from a different classification run on different imagery. Six months later someone in a parliamentary briefing asks why the two don't agree, and nobody owns a clean answer because nobody built a crosswalk between the two legends in the first place. The disagreement is a procurement artifact: two vendors, two schemas, no shared source.
Building a legend that crosswalks instead of competes
The fix is defining your national classification so each of its classes rolls up cleanly into a CORINE Level 3 code, the way a detailed crop taxonomy should nest under "arable land" without ambiguity. Done properly, one classification run on one year's imagery produces your domestic layer at whatever resolution your departments need, and a CORINE-compatible aggregation falls out of it rather than getting commissioned separately. You stop reconciling two independently built products and start publishing one layer with two legitimate views onto it.
This is also where an annual cadence earns its keep against CORINE's five-year refresh. If your national legend updates every year and crosswalks to CLC codes by construction, your five-yearly CORINE submission becomes an export of whatever year lands on the reporting calendar, not a separate commissioning exercise with its own classification contract and its own chance to drift from the domestic product.
Land Use Mapping is built around that idea: one classified layer per year, delta-checked against the previous edition, that your statistics office, environmental reporting team, and planning department can all cite from the same source, with a legend structured to roll up into CORINE-style reporting rather than fight it.
If reconciling two land cover products every reporting cycle is starting to eat more analyst time than the classification itself, it may be worth seeing what a single annual layer built for both uses looks like.