Sampling gaps

Each row is a place with lab samples, each column a month. A hatched cell had none. The months after September 2026 are not in this record yet.

Three seasons side by side. Each starts from an example to change, not a suggestion. Add spots from the map or the list above, or start a plan from a group of spots. The figures are arithmetic on the samples already taken; none says which season would teach the model more.

Beside these figures: the one rule for choosing samples that was tested, an information score, failed its test against equal coverage in a simulation, and no rule here has been tried in real water. Choosing places and months by the gaps they fill is a way to talk about a season, not a shown improvement. The simulation.

• No rule has been shown to work. The information policy failed its test against equal coverage (58.5% of simulated worlds, 60% needed). Nobody has sampled to any rule here and compared the result in real water.
• No samples is not clean water. A spot without samples is one the model knows little about. It says nothing about whether its water is better or worse.
• The samples are the Environment Agency's routine ones. They were taken on days fixed in advance, at bathing waters, from May to September. They show where the model's data are thin, not where the water is worst.
• “After rain” is archived forecast rain. More than 10 mm in the 48 hours before the sample, from the forecast as it was issued, not a rain gauge. Another line would give other counts.
• The rain share is a rough guide. It pools the bathing waters in an area. A spot in the same area can be wetter or drier, and a season can be unlike 2024–2026.
• The costs are yours. Both prices are assumptions you set, not a quote. Kit, training, delivery to a laboratory in time and staff on call after rain are not counted.
• A local group's samples are not in the model's data. Waterhead's samples are shown here for coverage. They have their own page and are not compared with the forecast yet.
• Official advice comes first. Environment Agency notices and signs at the water are on each spot page of the live site.

1. Does sampling by a stated rule teach the model more than a group's usual schedule?

   Useful if: the randomised week-level field study in the planner's record, with a pilot season first and its pass rule written down before the first sample. Unconvincing if: a rerun of the simulation, or a reanalysis of past EA samples, which no rule chose.

2. What do samples between October and April show?

   Useful if: a group that samples out of season shares dated results with the laboratory's method. Unconvincing if: a handful of winter samples is read as a winter rate.

3. How wet is a spot with no bathing water nearby on a fixed schedule?

   Useful if: archived forecast rain at each unsampled spot on a stated weekly day, 2024–2026, replacing the area share. Unconvincing if: it uses only the days a sample happened to be taken.

Back to the forecasts · The sampling planner · Research & data