Canada Data Curation¶

This document explains how to process and standardize the raw datasets for the Canada simulation.

Raw Data Sources¶

• data/canada/opencellid/302.csv: Cell locations for MCC 302.
  • Source: OpenCellID. 58,979 records, of which Telus holds 20,077 (302-220), Rogers 18,303 (302-720), Bell 11,237 (302-610), Freedom 4,899 (302-490) and Videotron 2,296 (302-500).
• data/canada/agent-unprocessed-raw-datasets/lcsd000b21a_e.*: Census subdivision boundaries, 2021 Census cartographic boundary file, projected in NAD83 Statistics Canada Lambert (EPSG:3347).
  • Source: Statistics Canada, 2021 Census boundary files.
• data/canada/agent-unprocessed-raw-datasets/98100002.csv: Population and dwelling counts by census subdivision.
  • Source: Statistics Canada, table 98-10-0002.

Census subdivisions are Canada's municipal level and play the role that municipios play in Spain and counties play in the United States.

Processing Script¶

data/processing_scripts/standardize_canada.py¶

1. Loads the population table and keeps census subdivision rows, identified by the A0005 schema marker inside the DGUID.
2. Reads the boundary shapefile with latin-1 encoding, since the attribute table carries accented French place names.
3. Reprojects every polygon from EPSG:3347 to WGS84 and simplifies it to a 200 m tolerance.
4. Joins geometry to population on DGUID, which is the only key present in both files, and drops subdivisions with no population.
5. Outputs municipalities.csv (4,830 subdivisions, 36,991,981 inhabitants, matching the published 2021 national count exactly), regions.geojson and cities.csv.

Official Site Data¶

Canada does have an official per-site registry, and it is the one to use.

• data/canada/ised/302.csv: 43,201 distinct carrier sites, built from the Terrestrial Spectrum Licence Site Data Extract.
  • Source: ISED, Download SMS data, file Site_Data_Extract_FX.zip, refreshed monthly. Cellular spectrum in Canada is licensed by area rather than per transmitter, but licensees must still declare the sites operated under those licences, and this extract is the result. Per carrier: Telus 14,974, Rogers 12,403 (including Fido), Bell Mobility 10,643, Videotron 4,030, SaskTel 1,151.

Do not use TAFL_LandMobile-LTAF_MobileTerrestre.zip for this. Despite the name it covers private land mobile radio and contains almost no carrier cell sites.

data/processing_scripts/standardize_canada_ised.py¶

One row of the extract is a licence, frequency and site combination, so the 1,772,480 rows collapse to 43,201 sites once deduplicated on rounded coordinates per carrier. Licensee names are mapped to MNCs, with the two Telus filing names merged and Fido folded into Rogers, whose network it shares.

Coverage: Why the Official Field Matters Here¶

Population living more than 10 km from any recorded site:

Two separate effects, worth keeping apart:

• Crowdsourcing gap. At network level the country is well covered, 4.17 % beyond 10 km, but OpenCellID reports 14.69 %. That difference is missing contributions, not missing infrastructure.
• Single-operator footprint. Taking one carrier's declared sites leaves 23 % of the population beyond 10 km in both sources, because Bell and Telus share radio access and each files only its own sites. A single-operator Canadian field therefore understates the coverage a subscriber actually experiences.

Run canada_ised when the absolute handover rate has to reflect real sites, and the all-carrier union when the question is national coverage rather than one operator's.

Scenario Parameters¶

julia --project main.jl national canada