1
Choose a catchment and dates
1 Jan 2024 to 30 Sep 2026
Catchment
Dates
Layers on screen
2
One timeline
tap a day to inspect it3
The chosen day
map and readings for the day4
The evidence rows
exactly what the timeline drawsShow the rows
5
What it cannot tell you
- Lining things up is not cause. A spill before a high sample does not show the spill caused it. Farms, birds, dogs, misconnected drains and unmonitored outfalls carry bacteria too.
- No reading is not a zero. Hatched stretches have no record. The EA has published 2026 spill records to the end of August, but they are not imported into this workspace yet, so 2026 shows no spills because nothing is loaded, not because nothing spilled.
- An overflow with no events in the file may have spilled with a broken or missing monitor. The annual return's own count is the check.
- “Would have flagged” is hindsight. It is the model refitted without that year's samples, run on archived rain. The site never issued it.
- Weight is a factor from 0 to 1. It ranks how strongly the model counts an overflow at a spot. It is not the share of bacteria that arrives.
- The satellite index is not a toxin measurement. NDCI rises with chlorophyll and floating plant material. It says nothing about blue-green algae toxins or about safe water.
- Rain is a reanalysis, not a gauge. ERA5-Land cells are about 9 km across. A storm on one fell can be missed or spread out.
- Official advice comes first. Environment Agency notices and signs at the water are on each spot page of the live site.
6
Questions worth asking next
proposed, not startedDo high samples at Ilkley follow spills at the three nearest overflows more than spills further up?
Useful if: a pre-stated comparison on 2027 samples, with the 2024–25 pattern as the stated hypothesis. Unconvincing if: it rests on the same 2024–25 samples this page shows.
Do Waterhead's weekly samples track Ambleside's treatment works spills?
Useful if: WET's 2026 sheet arrives and the EA's 2026 spill records (published to August) are imported. Unconvincing if: one wet week drives the whole pattern.
Would a gauge or radar rain layer change what a day looks like?
Useful if: cached radar (docs/RADAR-RAIN-CHECK-2026-10.md) is added and differs from ERA5-Land on the days that matter. Unconvincing if: it differs only on dry days.
Which other catchments have enough layers to be worth a workspace?
Useful if: a count of catchments with samples, spill records and rain in the same window. Unconvincing if: it picks catchments by how interesting their high samples look.