[EMS] Update Norway counties

[jsanz]

EMS has outdated counties for Norway that require a new version to be published. The current source SPARQL query works with Sophox (no need for a UNION anymore) and provides an updated list covering the country

#defaultView:MapRegions
SELECT
(if(bound(?id2),?id2,?id1) as ?id)
?iso_3166_2 ?label_en ?label_nn ?label_nb
WHERE {
  SERVICE <https://query.wikidata.org/sparql> {
  # Using nested query to ensure there is only one ?id2 value
  {SELECT
  ?id1
  (SAMPLE(?id2) as ?id2)
  (SAMPLE(?iso_3166_2) as ?iso_3166_2)
  (SAMPLE(?label_en) as ?label_en)
  (SAMPLE(?label_nn) as ?label_nn)
  (SAMPLE(?label_nb) as ?label_nb)
  WHERE {{
    # List of regions, whose sub-regions we want.
    VALUES ?entity { wd:Q20 }

    # P150 = "contains administrative territorial entity"
    ?entity wdt:P150 ?id1 .
    ?id1 wdt:P300 ?iso_3166_2

    OPTIONAL { ?id1 rdfs:label ?label_en . FILTER(LANG(?label_en) = "en") }
    OPTIONAL { ?id1 rdfs:label ?label_nn . FILTER(LANG(?label_nn) = "nn") }
    OPTIONAL { ?id1 rdfs:label ?label_nb . FILTER(LANG(?label_nb) = "nb") }
  }}
  # remove possible ID duplicates
  GROUP BY ?id1}
}
}

But this needs further investigation because the table from Wikidata/Sophox returns the following (simplified) table

iso_3166_2	label_en
NO-03	Oslo
NO-11	Rogaland
NO-15	Møre og Romsdal
NO-18	Nordland
NO-21	Svalbard
NO-22	Jan Mayen
NO-31	Østfold
NO-32	Akershus
NO-33	Buskerud
NO-34	Innlandet
NO-39	Vestfold
NO-40	Telemark
NO-42	Agder
NO-46	Vestland
NO-50	Trøndelag
NO-55	Troms
NO-56	Finnmark

While the ISO website returns

code	name
NO-03	Oslo
NO-11	Rogaland
NO-15	Møre og Romsdal
NO-18	Nordland
NO-21	Svalbard
NO-22	Jan Mayen
NO-30	Viken
NO-34	Innlandet
NO-38	Vestfold og Telemark
NO-42	Agder
NO-46	Vestland
NO-50	Trøndelag
NO-54	Troms og Finnmark

Being the differences:

• ISO 30 (Viken) was split in 2024 into 31, 32, and 33
• ISO 38 (Vestfold og Telemark) was split in 2024 into 39 and 40
• ISO 54 (Troms og Finnmark) was split into 55 and 56

This shows that Wikidata provides the latest ISO codes, but they are not yet approved in the ISO website.

Given that codes are not overwritten, only replaced, it would be possible to keep all codes, having then overlapping geometries, but ensuring clients with different datasets referring to different points in time for this dataset still find their geometries. In any case this needs review to ensure it does not break anything.

[jsanz]

Investigation: GeoIP database subdivision codes for Norway

I investigated which ISO 3166-2 codes the MaxMind GeoIP database currently uses for Norwegian counties, since EMS boundaries need to match GeoIP lookups.

Finding: MaxMind reverted to the pre-2020 codes in March 2024

In the March 2024 GeoNames monthly diff report, MaxMind states:

In 2020, Norway merged a number of their subdivisions, but this proved unpopular and they have now unmerged these subdivisions. We have used the ISO subdivision codes for these subdivisions from before the 2020 change. The ISO subdivision codes we will be using for these subdivisions are reflected in this diff report.

The diff report CSV confirms the transitions. Filtering for Norway entries:

curl -sL "https://dev.maxmind.com/csv/GeoNames-Monthly-Diff-Report-March-2024.csv" | grep ",NO,"

The old_subdivision_codes → new_subdivision_codes column shows:

Old code (2020 merged)	Reverted to (pre-2020)	Example cities
30 (Viken)	01 (Østfold)	Fredrikstad, Moss
30 (Viken)	02 (Akershus)	Lillestrøm, Sandvika
30 (Viken)	06 (Buskerud)	Drammen, Hønefoss
38 (Vestfold og Telemark)	07 (Vestfold)	Tønsberg, Larvik
38 (Vestfold og Telemark)	08 (Telemark)	Skien, Porsgrunn
54 (Troms og Finnmark)	19 (Troms)	Sjøvegan
54 (Troms og Finnmark)	20 (Finnmark)	Hammerfest

No Norway-related subdivision changes appear in the 2025 or 2026 release notes.

Implications for this issue

There are now three sets of codes in play for Norway:

Era	Codes	Used by
Pre-2020 (original)	NO-01 through NO-20 + NO-50	MaxMind GeoIP (current), EMS (current norway_counties_v1)
2020 merge	NO-30, NO-38, NO-42, NO-46, NO-50, NO-54	ISO 3166-2 (official, not yet updated)
2024 split	NO-31, NO-32, NO-33, NO-39, NO-40, NO-55, NO-56	Wikidata (current)

Key points:

• The current EMS data already matches what GeoIP returns (pre-2020 codes).
• MaxMind chose the pre-2020 codes, not the new 2024 codes (NO-31/32/33 etc.) that Wikidata returns.
• If the SPARQL query is updated to pull from current Wikidata, it would produce 2024 codes that do not match GeoIP, breaking the lookup alignment.
• Any update should keep the pre-2020 codes as primary identifiers to stay in sync with GeoIP. New 2024 codes could be added as supplementary data if needed for forward compatibility.

[jsanz]

Elasticsearch GeoIP validation

Scanned 2,560 Norwegian IPs via the ES GeoIP ingest processor (_simulate API, GeoLite2-City.mmdb on ES 9.3.0). Found exactly 15 unique region_iso_code values — a hybrid of pre-2020 and 2020 codes:

• Pre-2020 (split back): NO-01, NO-02, NO-06, NO-07, NO-08, NO-19, NO-20
• 2020 (stayed merged): NO-34 Innlandet, NO-42 Agder, NO-46 Vestland, NO-50 Trøndelag
• Unchanged: NO-03, NO-11, NO-15, NO-18

Not present: NO-04/05/09/10/12/14 (pre-2020 merged), NO-30/38/54 (2020 split-back), NO-31–33/39–40/55–56 (2024 Wikidata).

Current EMS data has 6 codes GeoIP no longer returns. To align: merge NO-04+05→NO-34, NO-09+10→NO-42, NO-12+14→NO-46.

Full details with reproduction steps: sources/no/GEOIP_INVESTIGATION.md (to be included in a follow-up PR).