How we know
The rules that apply to a storm overflow, how each spill is classified, when we flag a problem, and the documents behind every figure.
Method rev 7b59571 · every figure here is reproducible against this version · read the method ↗
Part 1 · The rules that apply
When a storm overflow may lawfully discharge
What is in the permit, and what is not
When an overflow gets investigated
Part 2 · How we classify each spill
Dry spill
A spill counts as dry when the nearest EA rain gauge recorded ≤ 0.25 mm on the spill day and the day before. A discharge with no rain to excuse it is presumptively non-compliant — an investigate signal, not proof of an offence.
Method ↗Spilled before its works
When an upstream overflow spills on a day its own treatment works did not, the works still had capacity — so the problem is more likely local (a blockage on that branch) than catchment-wide. We flag these as pre-works spills.
Method ↗Why very short spills are hidden
Monitors record in short intervals, and a single-interval blip often reflects sensor twitch rather than a real discharge. By default we exclude spills under 15 minutes so counts reflect meaningful events.
When a feed goes quiet
Live status comes from the water company's feed, which we poll hourly. A monitor that has gone quiet is not proof that nothing is spilling — it means nobody knows, so a quiet feed is shown as unknown, never as an all-clear.
Part 3 · When we flag a problem
| Problem | Fires at | Weight |
|---|---|---|
| High spill frequency | 400+ spills since 2020 | 4 (5 if 800+) |
| Spilling for very long periods | 500+ hours in the last full year | 4 (5 if 900+) |
| Dry spilling | 5+ dry spills since 2020 | 3 (5 if 15+) |
| Spills before its works | 4+ events since 2020 | 3 (5 if 12+) |
| Feed unreliable | no reading for 12+ hours | 2 |
Confidence in a dry classification
Each dry spill is scored High / Medium by how solid the classification is — how close the rain gauge sits, how wide the antecedent-dry window is, and how reliably the monitor was reporting. The score is shown on every event's evidence dossier.
What we cannot see
Misconnections, effluent quality from a normally-running works, flow transferred between works, maintenance history, and periods when a monitor was offline. Every count is a floor, not a total.
Thresholds are calibrated to the River Dart's 45 tracked overflows; an overflow's severity is the sum of its weights. Capacity verdicts use a 100% band (over the permitted flow) and a 95% band (at the limit).
The data is public — the links are not
Every dataset behind this site is already in the public domain: the Environment Agency's spill records, the WINEP measures, the permits, the rainfall. What is not published is the connective tissue — which measure addresses which overflow, which permit governs which outlet, which problem has an action against it and which is a gap. We assemble those links by hand, and where a link is missing we say so. Making these connections — so the public can see at a glance whether a failing overflow is being fixed — should be the job of the regulator, not of a volunteer group reconciling spreadsheets.
Sources & method documents
The dry-spill method is pinned to commit 7b59571 to match the method version; the rest track the main branch. Capacity and permit-derived figures are indicative estimates — no figure here should be quoted in a consultation response or funding bid without opening the source. Spill data: Environment Agency EDM returns; live status: water-company near-real-time feeds. Not a substitute for official advice.