METHODOLOGY · PRIMARY SOURCES · REPRODUCIBLE

Canadian Climate Data — Methodology & Sources

How we build our Canadian climate datasets. Every frost date, degree-day and frost-risk figure on this site traces to Environment and Climate Change Canada's published climate normals, read from a named long-record station by its Climate ID — nothing modelled or estimated. This page documents the sources, the stations, the reference periods and the integrity rules, so you can verify or reproduce any number.

The Dataset Family

Each open dataset, what it measures, its ECCC source product and reference period. Counts are read live from the published CSVs.

Station Registry — the 19 Long-Record Stations

The reproducibility backbone. Every measured dataset is keyed to these ECCC long-record stations by Climate ID. To check any figure, pull that Climate ID from ECCC's climate-normals results page. The right-hand column shows which datasets draw on each station.

Registry compiled live from the frost-free-shift, growing-degree-days and frost-risk CSVs. The frost-dates and gardening-data hubs cover a wider single-period city set (38 cities) drawn from the same ECCC normals.

Integrity Rules

• Published data only. We report ECCC's published climate-normal values. Nothing is modelled, estimated or interpolated. The single exception is cultivar hardiness, which is breeder/nursery/arboretum-rated (not measured) and is flagged as such on its page and cross-checked against our own growing guides.
• Same-station by Climate ID. Multi-period datasets read the same physical station in both reference periods, matched by Climate ID — a true like-for-like comparison, not an artefact of a moved station.
• Flag and skip, never substitute. If the same station lacks a published value in a period, the city is excluded and the exclusion is disclosed — e.g. Halifax is left out of the frost-free-shift study (no published 1991–2020 mean) and Thunder Bay out of the growing-degree-days study (its normals station publishes no degree-day section). We never quietly swap in a nearby station.
• Completeness codes disclosed. ECCC assigns each normal a completeness code (A = the WMO "3-and-5" rule; B ≥ 25 years; C ≥ 20; D ≥ 15). Where a value rests on fewer complete years, the code is carried through to the dataset rather than hidden.
• Live-computed. Every aggregate on our data pages (counts, means, warmest/coolest, rose/fell) is computed from the downloadable CSV on each page load — nothing is hand-typed, so a page can never disagree with its source data.

How to Reproduce Any Figure

1. Find the station's Climate ID in the registry above (or in any dataset's CSV).
2. Open ECCC's Canadian Climate Normals. The 1981–2010 and 1991–2020 results pages both accept a Climate ID directly, so you can call up the same station for either period.
3. Read the value in question — the Frost section (last/first frost, frost-free period, and the 10–90% probability rows), or the Degree-Days section (Above 5 °C / Above 10 °C, annual).
4. Compare against our CSV. For a two-period dataset, repeat for both the 1981–2010 and 1991–2020 normal at the same Climate ID.

Primary Sources

• Environment and Climate Change Canada — Canadian Climate Normals (1981–2010 and 1991–2020): climate.weather.gc.ca/climate_normals. The source for all frost, frost-free-period, degree-day and frost-probability figures.
• Natural Resources Canada — Plant Hardiness Zones: planthardiness.gc.ca. The source for the NRCan hardiness zones we cite. NRCan's official zone maps are the 1961–1990 and 1981–2010 models; we cite the NRCan zone and do not compute our own.
• World Meteorological Organization — the 30-year climate-normal standard that defines a reference period.

The Datasets