How WaterWatch works

Three views you switch between at the top of the map: Open water (rivers, lakes and coast), Bathing (designated swimming waters and community-added spots), and Drinking (public water supplies). Each has its own status scale. You can add reports on Open water and Bathing; Drinking is view-only, showing official monitoring.

Colours depend on the view. Open water uses the official EPA Water Framework Directive scale: High, Good, Moderate, Poor and Bad. Bathing uses the EU bathing-water scale: Excellent, Good, Sufficient and Poor. Grey means no official rating or no data yet, including a community spot with no citizen reports so far.

A beach or swim spot that is not one of the EPA-designated bathing waters. It is shown with no official rating, carrying only citizen reports until enough of them agree. If your usual spot is not listed, you can add it by name when you make an observation.

Two kinds of source. Official monitoring, from the EPA in the Republic of Ireland and DAERA in Northern Ireland. And reports from members of the public who have seen the water in person. Both are shown on the same map.

Official monitoring is the starting point wherever it exists. Public reports are added to it. Official data can raise or confirm a concern at any time. A public concern only changes the colour of a water body once more than one person has observed it with supporting evidence. A single public report never repaints a named river on its own.

The Drinking view shows only the public water sources the EPA monitors: designated abstraction rivers, lakes and groundwater, not every river. Private wells (about 12% of supplies) are unmonitored and not shown. A source shows as “Not flagged” in a neutral slate when it is not on the EPA’s Remedial Action List. That is a regulatory status, not a verified water-quality test, so we do not paint it a confident green. Supplies on the list show as At risk or Action required. Northern Ireland sources show grey, because compliance is not published there as open data.

An observation is a note you add about water you have seen in person, at a place and time. It can be as simple as what the water looked like, or it can include a photo or a test result.

On the map, tap or click a place to open a card showing what is known about that location. To start your own, use the Add an observation option shown, in that card or at the top of the screen. For a beach, tap it and choose "Add an observation here", or use Add an observation and then Set location on map to drop the pin yourself. Choose what you saw, add a photo if you have one, and submit. You can report with or without an account. On a touchscreen, you can also press and hold a spot to begin an observation there.

Anyone. You can report anonymously or with an account, and both count. Reports can be added on Open water and Bathing, but not on Drinking water, which is shown for information only.

No. A single report shows as a pin and is marked as needing confirmation. It does not change the colour by itself. This stops one mistaken or bad faith report from labelling a named river.

At least two independent reporters, and the reports need real supporting evidence, not just two reports with nothing to back them up. So two people who photograph a pollution problem will change the reading. Two empty reports will not.

When there is a report but not yet enough to confirm it, the water body keeps its current colour and shows a short note, for example "1 of 2, needs confirmation", along with the pin. When a second independent report with evidence arrives, the reading updates. This works the same for good reports and bad reports.

By how much supporting evidence it carries. A short written note counts the least. Adding a photo, structured details, or a verified test kit result each raises how much the report counts.

Several things together. A reading needs more than one independent person. Reports need supporting evidence to count, so a pile of empty reports achieves nothing. Repeat reports from the same person at the same place on the same day count once. And anyone can flag a report they believe is fake, which sends it for review.

Yes. You can make a positive observation as well as a negative one.

No. Positive and negative observations are never averaged against each other. A run of "all fine" observations cannot bury a confirmed problem. This is deliberate, so a real concern cannot be hidden behind positive observations.

Yes. A report counts for most when it is recent, and its weight fades gradually, halving about every 30 days. Older reports are not thrown away straight away, they count for less.

It depends on the water. For designated bathing waters, where the agency publishes each sample date, the card shows when the water was last tested, for example "Last tested 21 July 2026". For rivers, lakes and coastal waters, the official ecological status is a multi-year survey with no single test date, and drinking-water supplies publish no test dates as open data — so there the card shows when WaterWatch last checked the reading, for example "Last checked 1 May 2024", and marks it when that is more than three years ago so you can see it may not reflect the water today. When newer confirmed public reports disagree with an old official reading, the app shows that the public reports are updating a dated official assessment.

Yes. On a bathing spot the app shows live conditions — weather, wind, rainfall and, on the coast, sea state and tide — from Met Éireann and the Marine Institute, so you can judge the day rather than only the season. Recent heavy rain is also one of the inputs to the WaterWatch estimate, because it separates a short lived event from an ongoing problem. Conditions are context, never a water-quality rating or a swim-safety verdict.

Because it is still the most authoritative information available for that water body. Rather than hide it or present it as current, the app shows it and shows its age.

No. Report pins are shifted by a short distance, about 250 metres, so a report never points to an exact spot such as a home.

No. You can report anonymously. The public view never shows a reporter's identity or the exact time of a report.

It is planned. The app is being built to accept readings from fixed monitoring devices in future, so live sensor data can sit alongside official and public information.

Yes. The information is open. Local authorities and other official bodies see the same map as everyone else, which can help them decide where to test.

I built WaterWatch after taking my children to nearby beaches and rivers where I could visibly see the water was not clean, and it made me question where we could or could not swim. I wanted a way to check a recent picture of the water before we set out, instead of finding out when we arrived. There is also a wider need for this. An interview on RTE Radio 1's Countrywide discussed the case for an app of exactly this kind.

It is a modelled estimate of the chance that bacterial levels at a bathing water exceed the EU limits today. It is trained on the published history of EPA and DAERA samples together with recent rainfall and characteristics of the site. It is an estimate, not a test result.

No. The official EPA or DAERA classification is always shown alongside the estimate and always takes precedence. The estimate is extra context for the days between official samples, not a substitute for them.

It is validated before it is shown: it must beat the naive assumption that today looks like the last sample, both on later seasons it was not trained on and on beaches it has never seen. Every estimate carries a confidence level, and each one shows what drove it. It will still be wrong sometimes, in both directions.

Because conditions change. Heavy rain after a clean sample can raise the estimated risk, and a calm dry spell after a bad sample can lower it. A recent direct measurement always outweighs the model when it points to higher risk.

Because they measure different spans of time. The official EU rating is calculated from several years of samples, so it describes a beach's usual condition. The estimate is a nowcast for today. Short-term conditions like recent rainfall can raise the estimated risk for a day or two even at a beach whose long-term rating is better, so the two can point in different directions without either being wrong. The official EPA or DAERA classification remains authoritative, and for water-safety guidance follow Water Safety Ireland.

An estimate appears only where the model is in coverage for the spot, meaning a mapped, designated bathing water with enough recent data, and its live checks are passing. A data-poor spot, a community-added spot, or one the model does not yet cover reads "not available yet" instead of a guess. If the model as a whole stops beating the simple baseline it is withdrawn and refreshed rather than left to drift. Either way, no estimate never means the water is fine, and the official rating is always shown.

The estimate is tuned to a balance between catching real contamination and not raising too many false alarms, set from the published research on bathing-water nowcasting rather than picked by hand. It will still miss some events and flag some clean days, which is why it is only ever an estimate shown beside the official rating, never a verdict.

No. It is one more piece of information about water quality. It says nothing about currents, waves, tides or cold water, and it is not advice. Follow Water Safety Ireland, any local notices, and your own judgement.

How much rain fell in the days before today, taken over several windows from one day up to five. Rain washes bacteria off farmland and streets and can push storm water into rivers and the sea, so of everything the model looks at, this is the part most likely to change from one day to the next. It can push the estimate either way: a dry, calm run of days pushes it down.

The most recent official sample taken at that beach: how much bacteria it came back with, whether it was over the EU limit, and how many days ago it was taken. A recent clean sample pushes the estimate down, and it counts for less the older it gets. That fading is the gap the estimate exists to fill, because official sampling is roughly fortnightly in season.

The beach's own long run record: how often samples there have gone over the EU limits across the whole published history, and how much history there is to go on. It is the largest input at roughly three beaches in four, which is why an estimate can sit above the flag line on a dry, calm day. It is not the same thing as the last official sample, which is one day rather than years.

The fixed facts about the place: the land around it, how far the nearest storm water outfall is and how many sit within a few kilometres, whether it is coastal or inland, the usual rainfall for that area, and tide range where it applies. These barely change, so this input pushes the estimate by roughly the same amount every day at a given spot. The outfall figures are a distance proxy only. Ireland publishes no live data on sewage overflows, so nothing here says that anything was discharged.

The month, handled so that December sits beside January rather than eleven months from it. Bacterial levels follow a seasonal pattern, and this input carries that pattern. It is not about the bathing season calendar, which is a monitoring rule rather than something the water does.

No. The five are groups, and each one bundles several inputs, about fifteen in total. They include four separate rainfall windows, the land cover around the spot, the distance to the nearest storm water outfall and how many are nearby, tide range, the usual rainfall for the area, and the beach's own sampling history. Splitting all fifteen out would read as more open and would in fact be less true: the four rainfall windows move together, so dividing the credit between them gives four small numbers that mean nothing apart and everything together. Nothing is left out of the five groups. Every input the model uses sits in exactly one of them, and a new input has to be put in one before it can be used.

Because the two indicators are estimated by different kinds of model. The E. coli estimate comes from a single model, so a ranking of its inputs is that model's own reasoning and can be shown as it is. The intestinal enterococci estimate is an average of three models that usually disagree about which input mattered most, and at about half of beaches they even disagree about whether recent rainfall pushed the estimate up or down. Publishing one of those orders would be closer to a coin toss than to an explanation, so for that indicator the inputs are listed without a ranking. The inputs themselves are the same either way.