Eighty-five years of daily weather, kept the way a station keeper kept it — one line per day, high and low, rain and snow — and then played out loud. Every mark below is an observation somebody actually wrote down.
Three records, ten to twenty kilometres apart. They do not agree with each other, and that disagreement is half the story.
Each hairline is one day, drawn from that morning’s low to that afternoon’s high, inked by how far it sat from normal. Press play and the pen sweeps forward while you listen to it.
The picture and the sound are the same numbers. Higher pitch means a warmer day. The two voices you hear are the daytime high and the overnight low.
“Departure” means how far that day sat from what was normal for that date between 1951 and 1980. “Rank” counts every other time that same calendar date has come around at this station.
Computed live in your browser from the same numbers on the tape — nothing here was typed in by hand.
Across all three stations the overnight low has risen two to three times faster than the afternoon high. The days you notice have warmed a little. The nights you sleep through have warmed a lot.
A hot afternoon is obvious. A night that never cools off is the thing that actually changed — and it is the harder one on people, on power grids, and on anything that relies on the dark to recover.
Thirty-year averages, 1951–1980 against 1996–2025.
Every row is an average of thirty years of daily observations, so a single hot summer cannot move it much.
The outliers in this station’s file. Tap one and the pen jumps to that day.
Sources, handling, and the places where these numbers should not be trusted too far.