Methodology & data sources
Updated:
Every number on Atlas of Seasons comes from an open dataset or a public service. These pages list them by data domain — what we use each one for, how detailed it is, how often it is refreshed and what it cannot tell you.
Data domains
- Climate normalsMonthly temperatures, rainfall, rain days, sunshine, humidity and wind are long-term averages from gridded climate datasets, sampled at each place and corrected for altitude.
- Forecasts & outlooksForecasts come from numerical weather models served through Open-Meteo. Beyond about two weeks we switch from a day-by-day forecast to trends and seasonal anomalies applied to the local climate normals.
- Sea temperatureMonthly sea temperatures are a satellite-based climatology; today’s value and the coming days come from a marine forecast model.
- Nature modelsCherry-blossom and autumn-colour dates are estimated with temperature models calibrated on official Japanese observations; aurora chances combine the geomagnetic forecast with cloud cover and darkness; wildlife months come from verified observations.
- HolidaysPublic holidays come from an open holiday database, with corrections from official calendars; school holidays come from an open regional calendar for 36 countries, plus Türkiye.
- Crowds & pricesCrowd levels and price levels are estimates. We use the strongest statistic available for each place and fall back to weaker signals where nothing better exists.
- Safety advisoriesSafety notes summarise official travel advice from two governments. Practical facts such as plugs, voltage and driving side come from Wikidata; emergency numbers are checked by hand.
- Places, guides & photosThe list of places, their names in many languages, guide excerpts, attractions, photos, airports and maps all come from open, community-maintained sources.
Principles
- Open data only: every dataset we use can be checked by anyone, and its licence allows reuse. Licences that require attribution are credited on these pages and in the footer.
- Calculated, not copied: verdicts such as “best time to visit” are computed from the data with the same rules for every place, in every language.
- Quoted text stays quoted: excerpts from Wikipedia and Wikivoyage are shown as published (apart from formatting clean-up), with a link and their CC BY-SA licence, and only in their own language.
- Estimates are labelled: crowd levels, prices, nature forecasts and long-range outlooks are model estimates, and the pages say so.