How the environmental ranking works.
The same measures. Every place.
A consistent footprint
We use fixed GHSL 2025 urban-centre and dense-town polygons, with dated population estimates. A city proper in one country is not compared with a much larger administrative region elsewhere.
Two common pillars
Air quality combines modelled PM2.5 and surface ozone. Thermal conditions use the share of resident-hours with strong heat stress (UTCI ≥32°C) or moderate-or-worse cold stress (UTCI <0°C). The balanced lens gives each pillar 50%; the other lenses use one pillar.
Population-weighted conditions
Grid values are weighted by the residents within each place. Thermal thresholds are applied to each grid cell and hour before averaging. Temperatures and pollution concentrations are not divided by headcount. These are outdoor background estimates, not measured personal exposure.
Regular checks, complete months
The edition-month view compares conditions during the same complete calendar month. A rolling twelve-month view requires twelve verified months and remains unavailable until that history is ready. The native Mac batch checks for changed inputs overnight and resumes its saved queue after a restart. Its dated verification status is shown on the home page. Checking more often does not turn monthly source data into daily observations. Seasonal differences are expected.
Bands rather than false precision
Lower exposure receives a higher relative score. Scores are rounded into five-point steps, with shared ranks. Scores describe the entire qualified global cohort. Regional/type filters retain those ranks rather than calculating a new national percentile.
A comparison of conditions. Source grids are approximately 25–31 km. Nearby places can share the same background estimate. These bands are not medical, regulatory or overall best-place ratings.
Observed publication requires consistent GEOS-CF v2 analysis, ERA5-HEAT, source hashes, population/geography versions and at least 98% covered population and valid hours. These coverage thresholds are editorial safeguards. Incomplete inputs stay unranked. GEOS-CF v1/v2 are not spliced into a trend.