Why the threshold is local
A single national threshold measures the wrong thing. At 90F, Phoenix clears it for a third of the year in both periods and the change vanishes into a number that was already large; Seattle barely clears it in either and the change vanishes into a number that was already zero. Both readings are artefacts of the threshold. A local percentile asks the question that survives translation: how often does this place now reach what used to be a hot day here.
Why the count is annual
The threshold comes from summer, but the count runs across the whole year, because a day at that temperature in May or October is the change this exists to show. Restricting the count to June through August would define the lengthening season out of the measurement.
What the data is, and is not
ERA5 reanalysis, retrieved through the Open-Meteo archive. A reanalysis is a model reconstruction constrained by observations, on a grid of roughly 31 km. It is consistent across six decades and across cities, which station records are not: stations move, change instruments, and are surrounded by cities that grew around them. The cost of that consistency is that no single value here is a thermometer reading, and a coastal grid cell blends sea surface temperature with land.
Cities excluded
A number can be arithmetically correct and physically wrong. These cities are measured but not published, with the reason stated rather than dropped quietly:
- San Francisco: trends at -0.45F per decade (Mann-Kendall p=0.042), a significant cooling. Every independent surface record for this region warms over the same period, so this reanalysis cell is most likely blending sea surface temperature rather than measuring the city.
Last updated 2026-08-15.
Sources. ERA5 reanalysis via the Open-Meteo archive, daily maximum 2 m air temperature.