Denmark pilot: INSPIRE administrativeunit mask + CORINE cover #218

New issue

Open

opened 2026-08-24 17:05:13 +00:00 by art-bot · 0 comments

art-bot commented

2026-08-24 17:05:13 +00:00

Collaborator

Gated on #211 (the catalogue). Mask plus cover for Denmark, both from the anonymous
Klimadatastyrelsen INSPIRE route — no token needed for the pilot.

Mask: https://api.dataforsyningen.dk/ogcapi/features/au_inspire/collections/administrativeunit/items
→ 105 units; filter nationallevel = 1stOrder for the country polygon
(name_gn_spell_spellofna_text = "Danmark", nationalcode = DK). Verified live,
anonymous, CC BY 4.0. This is also the clean answer to the Greenland/Faroes scope
question — they are separate country codes, not Danish administrative units in this
dataset, so no filtering hack is needed. Say that in the config comment so a later reader
does not go looking.
Cover: CORINE CLC 2018 via the existing clms: provider, identity mapping (R116),
NUTS0 DK. Denmark's own lc_inspire/landcoverunit (576 118 polygons, anonymous,
CC BY 4.0) is far finer, but its classification is a URI scheme into the Danish
grunddatamodel that must be enumerated and argued class by class in the
conf/se/conf/pl style — that is the later upgrade, not the pilot.

Denmark-specific

CRS is normally EPSG:25832; the OGC API returns GeoJSON in EPSG:4326. Check what
arrives.
Denmark is an archipelago plus a peninsula: ~400 named islands. Expect a many-piece mask,
and expect Natural Earth's 1:10 m coastline to disagree with it substantially at that
scale. That disagreement is a conf/world.yaml matter, not a Danish one — see the
warning below and #186.
The publisher renamed: SDFI/SDFE → Klimadatastyrelsen. sdfe.dk, sdfi.dk and
kds.dk all still answer. Pin whichever host the metadata record itself gives.

The three things the Greece pilot got wrong — do not repeat them

The mask is the country's own outline and it does NOT drive the world carve. Build it
from the national boundary source and stop. conf/world.yaml carves Natural Earth with
Natural Earth's own admin-0 polygons and reads only the country code from
map_country_mask_0, never its geometry (R144). Do not extend the mask over water to
"fix" grey background land at a coast — Greece burned two rounds on exactly that, and the
cause was in conf/world.yaml and applied to every country (#184, #186).
Never verify the carve by intersecting background land with the mask. That check is
tautologically true — background_land is Natural Earth minus the carve — so it passes
for any mask at all, including an empty one. The carve's real checks are R144/R145/R146 on
conf/world.yaml; a country config does not need its own.
R34's reference is the country's LAND, not its mask (R143). Reference the land staging
table explicitly, even when the mask happens to be the same geometry, or the check
silently stops meaning anything the day that changes.

Deliverables

conf/<cc>/ per R17: base.yaml, downloads.yaml, staging.yaml, acceptance.yaml,
01-country_mask.yaml, 02-cover.yaml (build order is lexical), country: <cc> (R16).
A Makefile import-<cc> target chaining import-world (R107) and a fetch-<cc> path
through the importer's declared-downloads route (R10).
Acceptance tests in the same commit: R119 (mask partition populated, the
conf/fi/01-country_mask.yaml pattern), R34 (cover gapless against the land table,
1% budget), R22 (coarse-tier vocabulary equals the level-1 rollup tier), R120 (no NULL
cover_class), and a decoded tile (importer tiles --check).
Paste the make acceptance tally in the PR — not ad-hoc SQL (AGENTS.md).

Verify against the live database before committing: partitions at every zoom, row counts,
NULLs in the columns that matter, a decoded tile.

Gated on #211 (the catalogue). Mask plus cover for Denmark, both from the **anonymous** Klimadatastyrelsen INSPIRE route — no token needed for the pilot. - **Mask**: `https://api.dataforsyningen.dk/ogcapi/features/au_inspire/collections/administrativeunit/items` → 105 units; filter `nationallevel = 1stOrder` for the country polygon (`name_gn_spell_spellofna_text = "Danmark"`, `nationalcode = DK`). Verified live, anonymous, **CC BY 4.0**. This is also the clean answer to the Greenland/Faroes scope question — they are separate country codes, not Danish administrative units in this dataset, so no filtering hack is needed. Say that in the config comment so a later reader does not go looking. - **Cover**: CORINE CLC 2018 via the existing `clms:` provider, identity mapping (R116), NUTS0 `DK`. Denmark's own `lc_inspire/landcoverunit` (**576 118** polygons, anonymous, CC BY 4.0) is far finer, but its classification is a **URI scheme into the Danish grunddatamodel** that must be enumerated and argued class by class in the `conf/se`/`conf/pl` style — that is the later upgrade, not the pilot. ## Denmark-specific - CRS is normally **EPSG:25832**; the OGC API returns GeoJSON in EPSG:4326. Check what arrives. - Denmark is an archipelago plus a peninsula: ~400 named islands. Expect a many-piece mask, and expect Natural Earth's 1:10 m coastline to disagree with it substantially at that scale. **That disagreement is a `conf/world.yaml` matter, not a Danish one** — see the warning below and #186. - The publisher renamed: SDFI/SDFE → **Klimadatastyrelsen**. `sdfe.dk`, `sdfi.dk` and `kds.dk` all still answer. Pin whichever host the metadata record itself gives. ## The three things the Greece pilot got wrong — do not repeat them - **The mask is the country's own outline and it does NOT drive the world carve.** Build it from the national boundary source and stop. `conf/world.yaml` carves Natural Earth with Natural Earth's *own* admin-0 polygons and reads only the `country` code from `map_country_mask_0`, never its geometry (R144). Do not extend the mask over water to "fix" grey background land at a coast — Greece burned two rounds on exactly that, and the cause was in `conf/world.yaml` and applied to every country (#184, #186). - **Never verify the carve by intersecting background land with the mask.** That check is tautologically true — `background_land` is Natural Earth *minus* the carve — so it passes for any mask at all, including an empty one. The carve's real checks are R144/R145/R146 on `conf/world.yaml`; a country config does not need its own. - **R34's reference is the country's LAND, not its mask** (R143). Reference the land staging table explicitly, even when the mask happens to be the same geometry, or the check silently stops meaning anything the day that changes. ## Deliverables 1. `conf/<cc>/` per R17: `base.yaml`, `downloads.yaml`, `staging.yaml`, `acceptance.yaml`, `01-country_mask.yaml`, `02-cover.yaml` (build order is lexical), `country: <cc>` (R16). 2. A Makefile `import-<cc>` target chaining `import-world` (R107) and a `fetch-<cc>` path through the importer's declared-downloads route (R10). 3. Acceptance tests **in the same commit**: R119 (mask partition populated, the `conf/fi/01-country_mask.yaml` pattern), R34 (cover gapless against the *land* table, 1% budget), R22 (coarse-tier vocabulary equals the level-1 rollup tier), R120 (no NULL `cover_class`), and a decoded tile (`importer tiles --check`). 4. Paste the `make acceptance` tally in the PR — not ad-hoc SQL (AGENTS.md). Verify against the live database before committing: partitions at every zoom, row counts, NULLs in the columns that matter, a decoded tile.

art-bot self-assigned this

2026-08-24 17:05:13 +00:00

art-bot referenced this issue

2026-08-24 17:08:46 +00:00

Denmark roads: CVF (anonymous, named, coarse class) vs GeoDanmark Vejmidte (token, full class) #231

art-bot referenced this issue

2026-08-24 17:08:47 +00:00

Denmark rail and stations: Banedanmark's ArcGIS Hub (licence unasserted — R2 first) #232

art-bot referenced this issue

2026-08-24 17:09:36 +00:00

Denmark water and places: INSPIRE hydrography with flowdirection, and Danske Stednavne #233

art-bot referenced this issue