Coding and AI classes in Loughborough

The learner downloads 12,419 daily mean flows for Pillings Lock from the Environment Agency, covering the water years from October 1991 to September 2025, with no missing days. The line is the National River Flow Archive's Q10 figure of 21.8 cubic metres per second. Python marks each day as above or below the line. 1,268 days are above it, 10.2 per cent, just as a Q10 should give. The first trap is to call those 1,268 days "high-flow episodes". They are days, and most of them belong to the same few spells.

The second step is declustering. Rivers wobble around a line: a day just above, one just below, then above again. Treating each wobble as a new episode inflates the count, so the program merges runs that are closer together than a chosen gap. The answer moves from 334 all the way down to 88. None of these is wrong; each answers a slightly different question. The learner has to state the rule beside the number, every time.

The same idea, under the name debouncing, lives in every keyboard and game controller. A physical button bounces for a few milliseconds when pressed, and without a debounce rule one press registers as several. The learner writes one function, count_episodes(flows, line, gap), and tests it on a made-up series where the correct answer is known, then on a button trace. Finally they note the limits: 808 of the days are flagged suspect and 352 estimated by the Agency, and the archive warns that reservoirs and water recycling works change the river's flow substantially.

Asking how things were counted is a good reason for Loughborough teenagers to keep learning to code in 2026. The longer argument is in why teenagers should still learn to code in 2026.