How this registry is built
Splices join consecutive eras of the same location end-to-end: each calendar year is served by exactly one underlying ECCC station ID. Splices are only made after checking the sites are physically close (lat/lon from ECCC's station inventory) and — where the records overlap in time — measuring the day-by-day bias between them over the overlap. Where a meaningful bias was found (e.g. Saint John), the records are kept as separate locations instead.
Merges (marked “virtual”) combine two colocated stations day-by-day: each date's row is taken from the first source that has real data that day, so one station fills the other's gaps. Merged files keep each row's original station name, so provenance is visible per day.
Missing precipitation reconstruction: some older records (notably Toronto's 1840s) leave Total Precip blank while recording rain and snow separately. Where the column is blank, pages reconstruct it as rain + snow at ECCC's standard 10:1 snow-to-water ratio — the same formula ECCC used where all three columns are populated (≈2,990 days recovered registry-wide).
Gaps: the fetcher discards years for which ECCC returns only blank placeholder rows, so a missing year file means ECCC has no real data for that year. The coverage and gap lists below are computed live against the actual data files, but only at year granularity — a year counts as covered as soon as its file exists, even if most of its days are individually flagged “M” (missing) by ECCC. That blind spot is real: Mactaquac has a file for every year since 1976, so nothing below flags it, but from 2006 onward only 28–85% of its days actually have data (see its notes). Each analysis page applies its own day-level completeness rule on top of this (e.g. annual-extremes charts require ≥300 valid days per year) — this page doesn't re-derive that automatically, so a station can look “fine” here while several of its years are quietly excluded elsewhere.
A caution on the earliest records: 19th-century observers often under-recorded light winter snowfall, so cold-season statistics from a station's first decades deserve more skepticism than modern data. Notes below are curated by hand; newly added stations appear automatically but without notes until reviewed.