What's really in the water

Worth saying plainly, because most of this kind of site implies otherwise. We have no buoys deployed and no buoyage in the Sound. Nothing here is a live reading off our own sensor, and where a number comes from somebody else's feed we say so.

What we have is the water itself, which is the part most people skip. We know the shape of this estuary, the depth of its bed, what the tide does to it hour by hour and what the rivers push through it. That is enough to work out where a spill goes, and it is the thing you cannot get from a sensor on a post.

Sampling water is expensive and slow, so the question is never "test more", it is "test where". Run the hydrodynamics on a real spill and you get a prediction with an extent: this reach, these hours, roughly this strength. That is a testable claim, and it tells you exactly where a bottle is worth taking and when.

So the sequence is model first, then a targeted campaign aimed by it, then those results back into the model. Each round narrows the parts we are guessing at. The biggest of those is what actually leaves the pipe, which nobody measures, and which a handful of well-placed samples during a real discharge would pin down better than a year of routine monitoring somewhere convenient.

Buoys come at the end of that, not the start. A permanent sensor is worth deploying once you know which few hundred metres of water it should be sitting in, and the model is how you find out.

This is what we intend to measure, not what we are measuring. The first two are the ones that move fastest after a spill, and the last is the pair the Environment Agency actually grades beaches on.

It follows the water. So we take the Ordnance Survey tidal and surface-water shapes for the whole South West, lay a 25 metre grid over them, and trace each live discharge outward through water only. It cannot cross land. Narrow channels concentrate it, the way they do in reality. Strength falls off with the distance the water travels, not the distance a gull would fly.

Rivers carry it too. Contamination runs down the river network from inland outfalls, and where a dirty river meets the sea we hand it to the coastal model, so a spill miles upstream can still reach the beach at the mouth. How heavily a site counts is set by its own six years of spill history, not just what it did today.

This runs everywhere, on every spill, in seconds. It is a good deal better than a circle on a map. It is still an estimate, so we draw it as a fading wash rather than a hard edge, because a hard edge would claim more than we know.

For the Tamar we built something heavier. It is a hydrodynamic model of Plymouth Sound, the Hamoaze and the tidal river: about 49,000 cells carrying the real depth of the bed, driven by the measured tide at Devonport and the gauged flows of the Tamar, Tavy, Lynher and Plym. Then we put a real spill through it.

Watch the tide work. The plume runs down the deep channel on the ebb and comes back on the flood. At the shore by Devil's Point it was detectable two hours after the overflow opened, crossed the enterococci line twelve hours after that, and was back under it inside a day.

Run the same spill on a spring tide and it goes somewhere else entirely. One beach gets three times better, another gets twenty-six times worse. That is why a fixed list of at-risk spots is wrong for half of every month, and it is the argument for forecasting this rather than surveying it.

Two things we do that are easy to skip. We carry both indicator bacteria and report the stricter one, because enterococci dies off more slowly than E. coli and its share of the plume grows the further it travels, so leading with E. coli would understate how close the water sits to the line at the range people actually swim. And we read the answer at the shore, in shallow water, not out in the channel where the plume runs several times stronger.