SEEA land account vs land cover classification: what's the difference
Statistics offices get asked this more often than you'd think, usually by a planning department that's just been handed a SEEA land account table and wants to know why the numbers don't match the land cover map sitting on the GIS server. They're related products, but they're not the same thing, and conflating them is how you end up explaining a discrepancy that was never really a discrepancy.
What a land cover classification actually is
A land cover classification is a wall-to-wall thematic layer. Every pixel or polygon in the country gets assigned a class, usually something like built-up, cropland, forest, water, and bare ground, based on what's physically on the ground as read from multispectral imagery. It's a spatial product first. You can query it, overlay it, clip it to a basin or a municipality, and get an area total for any class in any boundary you draw.
A land cover classification carries no accounting period of its own. One edition shows the country on roughly one date, or a seasonal composite close to it. Comparing two editions produces a change layer, and a GIS team builds that comparison by hand, overlay by overlay, after the fact.
What a SEEA land account actually is
A SEEA land account is a statistical table, not a map. It follows the System of Environmental-Economic Accounting's asset account structure: opening stock at the start of the reference period, additions, reductions, and closing stock at the end, all broken down by land cover or land use category. The categories in SEEA's own classification (SEEA CF, cross-walked to IPCC land use categories for carbon reporting) don't always line up one-to-one with whatever scheme your mapping agency uses internally, which is exactly where the concordance table earns its keep.
The land account's job is to balance. Opening stock plus additions minus reductions should equal closing stock, for every category, every year. That's an accounting constraint, not a cartographic one, and it's the main reason a land account can't be produced straight off a single classification run. It needs at least two comparable editions, mapped to the same class definitions, so the area moved from cropland to built-up (say) can be counted as a flow rather than noise from a reclassification quirk.
Where the two meet
The classification is the source; the account is the derived statistic. A statistics office pulls area-by-class totals from each year's classified layer, applies the concordance table to the account's categories, and populates the stock and flow rows. If the mapping agency changes its classification scheme, updates its sensor, or shifts its minimum mapping unit between editions, that shows up in the account as a reduction or addition that has nothing to do with actual land conversion on the ground. This is the single most common cause of a SEEA land account that doesn't reconcile with what the planning department sees on the map, and it's a methods problem, not an imagery problem.
Why agencies struggle to keep both current
The friction is operational. The classification gets commissioned through one contract, the account gets compiled by a different team on a different schedule, and the two rarely use the same class definitions or the same edition cadence. A statistics office ends up reconciling two products that were never built to reconcile with each other in the first place.
The fix is upstream: one classified layer, produced on a fixed annual cadence, delta-checked edition to edition, with class definitions stable enough that cropland means the same thing this year as it did last year. Land Use Mapping produces exactly that kind of annual base layer, the one your environmental accounts, your statistics office, and your planning department can all cite as the same source rather than negotiating between three versions of "cropland."
If your SEEA land account keeps drifting from your land cover map at renewal time, that's worth a look.