It is a model, not a thermometer. ERA5 is a reanalysis: observations fused into a physical model on a
grid of roughly 25 km (9 km for the land variables). It is the best consistent record that exists back to 1940,
but the value for your street can differ from your point's grid cell — most of all in mountains and on coasts.
Rain is weaker than temperature. Reanalysis precipitation carries much larger error than temperature,
especially for single days and for convective storms. Treat the rainfall figure as an indication, the temperature ranking as solid.
Days are UTC. Daily maxima and minima are aggregated over the UTC day, which for far-eastern
and far-western longitudes is offset from the local calendar day.
The odds are counted, not modelled. "1 day in N" is the observed frequency inside two 30-year baselines
(1941–1970 and 1996–2025). Where a value never occurred in a baseline we say so, instead of extrapolating a number we cannot support.