Methodology deep-dives, data journalism, and the occasional opinion piece from the team building WaterWatch. No spin. No paywall. Just the data.
WaterWatch doesn't just track spills that already happened. It now models the conditions that make them likely. Antecedent wetness, catchment runoff, per-site logistic regression, and a metric that tells you how many millimetres of rain stand between a site and its next discharge.
Counting spill hours tells you how often a CSO discharged. It doesn't tell you whether the river actually noticed. We cross-referenced 77,000 discharge episodes against EA water quality samples — here's what we found, and how we turned it into a 0–100 impact score.
A 17-year-old's completely honest, only slightly embarrassing explanation of why WaterWatch has ads, where they live, and why this particular page has quite a lot of them. AWS credits expire. Servers don't pay for themselves. Here's the deal.
Reducing spill hours year-on-year sounds straightforward. It isn't. Rainfall, sensor gaps, and three-year baselines all complicate the picture. Here's our honest framework for reading improvement signals.
Every number on WaterWatch comes from Thames Water's own EDM sensor network. We don't estimate, extrapolate, or fill gaps. Here's exactly how our pipeline works — and where its limits honestly lie.