Where your frost dates
come from.
Edena’s calendar turns on two public UK datasets. We credit them here in full because the Open Government Licence requires it — and because the science behind your sowing dates deserves more than a tooltip.
© Crown Copyright 2026. Information provided by the Met Office.
- Dataset
- Met Office HadUK-Grid v1.3.1 · 1km gridded observations · 1991–2020 reference period.
- DOI
- 10.5285/f02cc6ddd92f45b18b9ab6ab544df7d9Hollis et al., 2025.
- Licence
- Open Government Licence v3.0 · commercial use and redistribution permitted under the attribution terms above.
Postcode centroids come from the ONS Postcode Directory (ONSPD), released quarterly by the Office for National Statistics and licensed, like HadUK-Grid, under Open Government Licence v3.0. Rural postcodes span several 1 km cells; we use the centroid (Met Office / UKCP standard practice) and carry a confidence flag when the spatial extent is unusually large.
UK-FDP — a named, cite-able algorithm.
For every 1 km grid cell we take 30 years of daily minimum air temperatures and pull out, year by year, the last spring frost and first autumn frost— a frost being tasmin < 0 °C, the Met Office’s air-frost definition. We then reduce those 30-year distributions to three percentiles: 10th, 50th (median), and 90th.
“Safe to plant out”is the 90th-percentile last-spring frost — a date beaten by frost only one year in ten. “Protect tender plants by”is the 10th-percentile first-autumn frost. The spread between them is your frost-free growing season.
Postcodes resolve to district centroids (SW1A, EH1, IV1…) at roughly ±1 km horizontal accuracy. Northern Ireland postcodes are transformed from Irish Grid to OSGB36 before lookup. Channel Islands and the Isle of Man fall outside HadUK-Grid’s domain and are flagged as uncovered.
Frost dates are based on 1 km gridded observations. Sheltered valleys, frost pockets, and hilltop sites may experience significantly different frost timing — the COLPEX experiment measured 9.6 °C valley-hilltop differences within a single 1 km cell.
A short microclimate questionnaire will refine the estimate for your plot. Until then, treat the 90th-percentile date as your safe limit and trust local knowledge above any algorithm.