Kyoto cherry blossom (sakura) full-bloom dates, 812–2015 CE
Yasuyuki Aono's compilation from Heian-era court diaries, temple records, and modern observation. 1,204 years; 827 observations.
Pre — typical default chart
Connect-the-dots line across all valid years. Year-to-year jitter dominates; the long-term signal disappears into the noise.
Post — Tufte treatment
Each gray dot is one historical observation; the dark line is a 30-year running median. Y-axis inverted so earlier blossoms (= warmer spring) appear higher. Light shaded bands behind the data mark major climate epochs (labels above). Three records are circled: the earliest and latest historical observations, and the most recent year in the file. The signal: the 30-year median sat near the pre-industrial baseline (≈April 15) for over a millennium, then dropped to ≈April 5 in the industrial era — ten days earlier, a phenological fingerprint of modern warming.
The pre version contains exactly the same data and is not "wrong." But anyone who reads it walks away thinking "blossom dates vary a lot." Anyone who reads the post version walks away thinking "the medieval Japanese saw the spring arrive on April 15; modern Japanese see it on April 5." Same dataset, opposite takeaway.
What the skill changed
Connected line → dot scatter + running median. Separates noise from signal; honors graphical integrity.
Heavy grid → three labeled dashed dates + one baseline. Each line has a meaning a reader can use.