Home › Methodology
How we build these numbers
Every figure on this site is traceable to a public, free, public-domain or openly-licensed dataset — no proprietary data, nothing behind a paywall. This page explains where each kind of number comes from and how we turn it into the charts and tables you see. The "Methodology & sources" panel on each city page names the exact station and source for that city.
Measured vs. modelled
The most important distinction on the site is whether a number was measured at a weather station or modelled.
- Temperature and precipitation are observed. For most cities the 1991–2020 climate normals are an official 30-year product from a national meteorological service; for the rest we compute the normals from the station's own daily observation record. Either way, a physical instrument recorded them.
- Cloud, humidity, wind and sunshine are modelled. These variables are not measured at most weather stations, so we use NASA POWER, NASA's satellite-and-reanalysis climatology — the standard global source where ground measurement does not exist.
- A small number of locations have no usable nearby station. For those, the temperature and precipitation normals are modelled from ERA5-Land reanalysis on a ~9 km grid. The city page's methodology panel states this plainly when it applies.
Sources
NOAA U.S. Climate Normals 1991–2020
The official 30-year reference for U.S. stations, published by NOAA's National Centers for Environmental Information — monthly temperature, precipitation and snowfall normals. The prior 1981–2010 product is also loaded; the difference between the two periods drives the warming-delta table.
ncei.noaa.gov →Japan Meteorological Agency (JMA)
Official 1991–2020 climate normals for Japan — monthly mean, maximum and minimum temperature and precipitation, from the JMA principal-station and AMeDAS observation networks. Used directly for Japanese cities served by a JMA station.
jma.go.jp →INMET — Instituto Nacional de Meteorologia (Brazil)
Official 1991–2020 climate normals for Brazil, from INMET's network of conventional climate stations. Used directly for Brazilian cities served by an INMET station.
portal.inmet.gov.br →SENAMHI — Servicio Nacional de Meteorología e Hidrología (Peru)
Official 1991–2020 climate normals for Peru. We use only SENAMHI's full-period stations — those with a complete 24–30 year record over the window; shorter-period station averages are excluded. SENAMHI publishes maximum and minimum temperature, and the mean is taken as their average.
senamhi.gob.pe →CONAGUA / SMN — Servicio Meteorológico Nacional (Mexico)
Official 1991–2020 climate normals for Mexico, published per station by the Servicio Meteorológico Nacional under CONAGUA. Used directly for Mexican cities served by an SMN station.
smn.conagua.gob.mx →WMO CLINO 1991–2020
The WMO Climatological Standard Normals — official 1991–2020 normals submitted by 141 national meteorological services and compiled by NOAA NCEI. They serve two roles here: an independent reference for cross-checking our other normals, and — for countries where official coverage is otherwise thin — a direct source of city normals.
ncei.noaa.gov →NOAA GHCN-Daily & GSOD
Global daily station observations — the Global Historical Climatology Network (GHCN-Daily) and the Global Summary of the Day (GSOD). Where a station has no official normal, we compute its 1991–2020 normal from these daily records; the city page reports how many years of data went into it. They also feed the records and anomaly pages.
ncei.noaa.gov →NASA POWER
Cloud cover, relative humidity, wind speed and solar radiation — a satellite-and-reanalysis climatology (MERRA-2 and related products). The standard global source for atmospheric variables that station networks do not measure.
power.larc.nasa.gov →ERA5-Land reanalysis
The ECMWF ERA5-Land reanalysis, a ~9 km global grid of modelled climate fields. Used as the temperature and precipitation fallback for locations with no usable nearby station.
ecmwf.int →Beck et al. Köppen-Geiger climate classification
The reference Köppen-Geiger maps (Beck et al. 2018, CC BY 4.0; and the 2023 V3 1901–2099 dataset used for the climate-drift pages). Each city's Köppen tag is computed from its 1991–2020 monthly normals using the standard Peel/Beck rule set.
gloh2o.org/koppen →USDA 2023 Plant Hardiness Zone Map
The official U.S. gardening reference — each zone defined by the average annual extreme minimum temperature. Hosted by the PRISM Climate Group at Oregon State University.
prism.oregonstate.edu →Methodology
- Cities, not stations. Every page is built around a city, addressed at
/climate/{country}/{slug}/. Each city resolves to one canonical weather station; the methodology panel on the page names that station and its distance from the city centre. - Climate normals. A climate normal is the 30-year reference average defined by the WMO — monthly mean temperature and monthly total precipitation over the 1991–2020 window. Where we compute a normal from daily station records rather than use a published product, a calendar month counts toward the average only when that year has adequate daily coverage with no extended gaps, and the normal is built only from the years that qualify. Each city page reports how many years went into its figures.
- Köppen-Geiger classification. Computed from each city's 1991–2020 monthly normals using the rule set of Peel, Finlayson & McMahon (2007), the standard reference also used by the Beck et al. maps.
- Warming delta. The 1991–2020 monthly normals minus the 1981–2010 normals, shown only when both periods have a complete temperature series.
- Climate twins. Euclidean distance over the 12-month average-temperature vector; a lower distance means a more similar climate, in units of roughly °F per month.
- Anomaly. A recent day's temperature minus the matching month's normal — a deviation signal computed from GHCN-Daily observations.
Underlying datasets are in the public domain or openly licensed; redistribution and commercial use are permitted, with attribution to the original providers where their licence requires it. The providers are credited above.