SwimSignal Research Lab · Exhibit 3 of 4 · exploratory Accuracy

How good are the forecasts?

When SwimSignal says the water is likely to be polluted, how often is it right, and what does it miss?

Loading the record…

1

Choose a bathing water

2

Forecast against sample, one bathing water

E. coli, May–September 2024–2026

…

3

Against the simple alternatives

all 32 bathing waters with an estimate
4

Spill forecasts, a year held out

United Utilities overflows, 2025
5

Beside the Environment Agency's own forecast

16–28 September 2026
6

What it cannot tell you

  • These are not the forecasts the site showed. The E. coli estimates were rebuilt afterwards, leaving out each year in turn, with the weather forecasts as archived. The site has logged its own issued estimates only since 15 September 2026.
  • A sample is one bottle at one time. A miss means the model and that bottle disagreed. It does not say which overflow, farm or bird was responsible.
  • Nothing after 30 September 2026 is shown. Several forecast changes are being tested on autumn data now. Their results are read on agreed dates and stay closed until then.
  • The spill test covers one company. It held out 2025 at United Utilities overflows. The other eight English companies' scores are on the Accuracy page, from live data.
  • Over 900 is a line, not a safety boundary. It is the inland "sufficient" threshold for E. coli per 100 ml. Water under it can still make someone ill.
  • Official advice comes first. Environment Agency notices and signs at the water are shown on each spot page.
7

Questions worth asking next

proposed, not started