Methodology
How your preferences become a score, and what that score does not mean.
How scoring works
You give each measure an ideal range, a tolerance and an importance. Every city is scored month by month against those preferences, then combined into a single 0–100 score.
- Inside your ideal range scores full marks.
- Outside it, the score falls along a smooth curve measured in units of your tolerance, so nudging a preference never reshuffles your results dramatically.
- Breaking a hard limit sorts a city below every city that passes, however well it scores elsewhere.
Annual averages are never compared. San Francisco and Lisbon have similar annual means and very different climates. Seasons follow each city’s own hemisphere, so your summer preferences meet December–February in Sydney and June–August in Barcelona.
What this cannot tell you
It is not a forecast. These are 1991–2020 averages describing a typical year, not next year.
Neighborhoods differ from cities. A few miles inland from the California coast can mean ten degrees warmer on a summer afternoon.
Counts of extreme days run low. They come from a coarser model grid that smooths daily peaks, so days above 90°F and below freezing are conservative — treat them as comparable between cities rather than as a local airport’s figure. This matters most if you set a hard limit on hot days.
Climate is one input among many. Nothing here speaks to cost, safety, jobs, healthcare or visas.
Data
6146 cities across 172 countries, on 1991–2020 climate normals.
Checked against weather stations. Since October 2026, average highs, lows and rainfall are calibrated to station-based climate normals on a 4 km grid and adjusted to each city’s elevation. Against nearby weather stations in about 1,600 cities, the typical monthly error in average highs is about 1°F. Rainy days, hot days and freezing nights are recounted from 30 years of daily weather adjusted to those averages, with light drizzle no longer counted as a rainy day. Sunshine is calibrated against official sunshine records from more than 700 weather stations worldwide; uncorrected, the model overstated it by about a quarter of the possible daylight.
Spotted something wrong? Email stormwatchcolin@gmail.com with the city and what you expected.