A homogenization-free air-temperature urban heat-island dataset for 1,117 cities from GHCN-Daily, 1941–2020

Abstract Station-based urban heat island (UHI) studies generally use homogenized records, and homogenization removes the gradual local warming a heat island produces, erasing the signal under study: it removes about 60% of the global UHI intensification since 1971–90 and 83% of it in North America. This record therefore builds UHI from GHCN-Daily air temperatures that passed daily quality control but no homogenization. A city’s UHI is its urban station minus the median of rural stations 40–150 km away, corrected for elevation, and reported for the daily mean, night and day. It covers 1,117 cities over 2001–2020 and 719 with long station histories, from an annual panel of 8,637 stations, with within-city panels at five-year epochs from 1975 and 20-year windows to 1941. Every product is issued with and without a 12 km rural-reference screen. Coverage is concentrated in North America, Europe and Asia; the other continents hold about 13%. Validation covers the reference radius, lapse rate, station siting, per-city uncertainty and replication from an independent archive.