Skip to content

Raw Country-Specific Data for Portugal

Portugal exists for the Iberia roaming scenario (phase 2 of the §7.4 roaming work): a Spanish home operator's subscribers roaming into a Portuguese visited network across a real land border, giving geometric border crossings and a realistic Home-Routed hairpin (Lisbon serving site → Madrid PSA).

Portugal's official sources mirror Spain's INE + IGN pair exactly — Portugal's national statistics institute is also called INE (Instituto Nacional de Estatística, ine.pt), and the boundary authority is the DGT (Direção-Geral do Território) via the CAOP (Carta Administrativa Oficial de Portugal).

  • opencellid/268.csv: Cell Tower (gNB) locations for Portugal (MCC 268), 22,627 cells.
  • Source: OpenCellID (MCC-268 extract, retrieved July 2026).
  • agent-unprocessed-raw-datasets/ine_censos2021.json: resident population per municipality, Censos 2021 definitive (indicator 0011609, "População residente por Local de residência à data dos Censos [2021]"). Municipality rows carry the official 4-digit DICO code.
  • Source: INE Portugal JSON API.
  • agent-unprocessed-raw-datasets/caop_municipios.json: official municipality polygons, CAOP2025 Continente (mainland only), WGS84 GeoJSON, keyed by the same DICO code (dtmn).
  • Source: DGT OGC API (dataset registered at dados.gov.pt).
  • agent-unprocessed-raw-datasets/PT.txt: GeoNames country dump (PT.zip, CC-BY-4.0) — used only for cities.csv map labels.

Helper Script

data/processing_scripts/standardize_portugal.py

Joins INE population to CAOP polygons on the DICO code (CAOP dtmn == INE geocod), computes centroids with shapely, and writes the simulator standard format:

  1. municipalities.csv — id,name,population,lat,lon; id = DICO, population = Censos 2021, centroid from the CAOP polygon. 278 mainland concelhos, population 9,855,909 (the exact Censos 2021 mainland figure). The islands cut (Madeira/Azores, mirroring Spain's Canary filter) falls out of the CAOP Continente join for free.
  2. regions.geojson — the CAOP polygons re-emitted with feature property id = DICO (the loader Int-normalizes ids on both sides, so the leading zero in e.g. 0101 is harmless). Verified: 278/278 municipalities load with polygons attached.
  3. cities.csv — 9 mainland cities above 100k population, from GeoNames.
source data/processing_scripts/.venv/bin/activate   # shapely needed
python3 data/processing_scripts/standardize_portugal.py

Operators (net ids verified against the OpenCellID 268 data)

Operator net id cells
Vodafone PT 1 7,695
MEO (Altice) 6 6,791 (+4,091 under net 2, MEO's ex-TMN allocation)
NOS 3 4,050

main.jl national portugal uses MEO (net 6) with 18 mainland distritos as the edge-UPF count (the Spain-provinces analogue): 6,506 mainland gNBs → 18 edge UPFs → 2 PSAs, all 278 municipalities polygon-matched.

Iberia scenario (phase 2 — IMPLEMENTED, main.jl iberia (runs/iberia.jl))

Two operator fields composed into one topology via DataLoading.compose_topologies (Movistar 52 edge/5 PSA + MEO 18 edge/2 PSA → 52,902 gNBs, 70 edge UPFs, 7 PSAs, 8,322 municipalities; operator tags on gNBs and PSAs). Agents place across both countries by merged Censos/INE population weights. An agent whose nearest gNB flips operator has geometrically crossed the border, which is the roaming trigger (no coverage-overlap machinery needed). 5G charges the border per RoamingConfig.border_semantics; 6G-RUPA renumbers into the visited DIF. Home-Routed path-stretch becomes measurable: serving PT edge UPF → pinned Spanish PSA.