Methodology

How Swimmable decides what to say about each bathing water.

What this map shows

Swimmable is a reporting layer over the data the regulators and water companies of Britain are required to publish. It does not measure water quality. It does not predict water quality. It surfaces what other authorities are saying — the Environment Agency, Natural Resources Wales, the Scottish Environment Protection Agency, and the eleven water and sewerage companies that operate Britain’s public discharge feeds — and aggregates those signals into a single map status for each designated bathing water.

The map covers bathing waters across England, Wales, and Scotland. These are the swimming locations that have been officially designated by their regulator, and they are a small slice of where Britons actually swim. The River Thames at Henley is not on the map; nor is most of the Lake District; nor are the thousands of wild swimming spots catalogued by the Outdoor Swimming Society. The absence of a place from this map does not mean it is unsafe, or that nobody swims there. It means there is no regulator-published water quality data to anchor a status decision.

Each spot on the map carries one of five states.

A green pin means safe to swim. The bathing water carries a favourable rolling classification from its regulator (Excellent or Good), a fresh recent sample backs that classification, and no concerning discharge activity has been recorded in the last 48 hours within 5 kilometres. Today the map shows spots in this state.

An amber pin means use caution. Something specific has flagged a concern: a sewage discharge in the last 48 hours, an active pollution risk forecast, or a rolling classification of Sufficient (the regulatory pass mark, but not a state we are willing to call safe). The map currently shows spots in caution.

A red pin means avoid swimming. The bathing water carries a Poor classification, has been formally closed by its regulator, or has an actively-discharging sewage outfall within 5 kilometres. spots are in this state.

A pale blue pin means monitored — no current concerns. This is the state for spots where the regulator’s rolling classification is favourable but no fresh sample has yet arrived. It is not a green pin, because we will not claim “safe to swim today” without current evidence; it is not a red pin, because nothing specific has flagged a concern. It is the middle state, and of Britain’s designated bathing waters sit here today. That number rises and falls with the sampling season.

A grey pin means status unknown. The data needed to produce a verdict isn’t currently available — typically because a spot is outside its regulator’s monitoring season, or because the regulator has not yet issued a classification for it. spots are in this state right now; from mid-September through mid-May, every Scottish spot is.

Across Britain at this moment, storm overflows are actively discharging sewage. The current count of active discharges varies hour by hour, much higher after rainfall, lower in dry weather. Most of them are inland sewage works far from any designated bathing water. designated bathing waters are currently within 5 kilometres of an active discharge — they appear in red or amber on the map above.

The map looks different in May from how it looks in July. In the weeks after the statutory English bathing-water season begins on 15 May, most English spots carry only samples from the previous autumn, and the map sits heavily on pale blue. By mid-July, as 2026 in-season samples flow through the publishing pipeline, the map becomes substantially greener. By mid-October the samples taper off again. SEPA monitors only between 15 May and 15 September; outside that window every Scottish spot reads unknown. This rhythm is a property of how British water-quality monitoring actually works, and the map makes it visible rather than hiding it.

Throughout, the principle that governs every editorial decision is that absence of fresh signal is not the same as fresh evidence of safety. A swimming spot that hasn’t been tested in six months is not “safe” because nothing bad has been reported. It is unmonitored, and the map says so.

Where the data comes from

The map draws on thirteen distinct sources of public data: three regulators and ten water and sewerage companies. (An eleventh water company, Welsh Water, is treated separately below because what it publishes is structurally different from the rest.) Each source is consulted independently; the map’s verdicts are aggregations of what each authority is saying about water it monitors or sewage it discharges.

The regulators. Three statutory bodies are responsible for classifying designated bathing waters in Britain.

The Environment Agency monitors designated bathing waters in England. of them are newly-designated for the 2026 season — including river bathing sites such as Pangbourne Meadow and Ham & Kingston on the Thames, Falcon Meadow on the Waveney, and Sandy Lane on the Dee at Chester — and carry no classification yet because the first sampling round is still to come. The EA publishes annual classifications (Excellent, Good, Sufficient, Poor, or Closed), short-term pollution risk forecasts during the bathing season, and the dates of its most recent samples. It publishes via a documented Linked Data API and is, of the three regulators, the most complete and consistent data publisher.

Natural Resources Wales monitors Wales’s designated bathing waters. It uses the same Defra infrastructure as the EA and publishes classifications and sample dates of equal quality. NRW publishes risk forecasts for only of its spots — about of Welsh bathing waters have the same active-pollution-warning capability that every classified English spot has.

The Scottish Environment Protection Agency monitors Scotland’s bathing waters during a statutory season that runs from 15 May to 15 September. For the rest of the year — two-thirds of every calendar year — SEPA does not sample. It publishes via server-rendered HTML pages rather than an API, does not publish short-term risk forecasts at all, and one of its designated bathing waters carries a published grid reference that cannot be parsed by anyone — Lunderston Bay, with a 9-digit grid reference that violates standard OS conventions. Until SEPA corrects its own publication, that location cannot appear on this map.

These three regulators implement the same European Bathing Water Directive heritage but produce structurally different qualities of public data. The asymmetry is not the map’s choice; it’s what the regulators publish.

The water companies. Ten privately-owned English water companies plus Scottish Water (publicly-owned) operate the wastewater infrastructure that discharges into Britain’s bathing waters. Under the Environment Act 2021, each is required to monitor and publish near-real-time information about every storm overflow they operate. The Act mandates publication; it does not standardise format or quality.

The result is that some water companies publish exceptionally — Southern Water publishes five and a half years of event-level history and is the only company in Britain to publicly flag which of its own events it considers genuine sewage discharges versus probable monitor false-trips. Thames Water publishes a documented first-party API exposing four years of event history. Wessex Water publishes about sixteen months. Yorkshire Water publishes about four. Anglian, Northumbrian, Severn Trent, South West, United Utilities — five major English water companies — publish only the current state of each overflow plus its single most recent event, with no historical depth available. Scottish Water publishes a structurally distinct feed exposing the current and previous event for each of its overflows.

Three operators — Welsh Water (Dŵr Cymru), Wessex Water and Yorkshire Water — publish, per overflow, a curated list of which bathing waters that overflow is operationally linked to. This is the operator’s own hydrological assessment, not a spatial inference, and the status engine treats it as the highest-priority proximity signal: ahead of the 5 kilometre Euclidean radius, ahead of the catchment-aware downstream check. Across the three operators the map captures roughly three hundred and sixty resolved (overflow, bathing-water) linkages. Two of Welsh Water’s linkages fall beyond a simple 5-kilometre radius — Penarth Kymin Storm Overflow to Jackson’s Bay Barry Island at 8.6 kilometres, and Rhoose Storm Overflow to Whitmore Bay at 5.5 kilometres — operator-curated examples of catchment effects a purely geometric proximity model would miss.

One derived layer on top of the water companies’ data. Alongside those eleven operator feeds the map ingests one additional public dataset: the downstream-impacted river reaches published by sewage-map, an independent project by Alex Lipp, Jonny Dawe, and Sudhir Balaji that powers sewagemap.co.uk. For each of ten English and Scottish water companies plus Welsh Water, sewage-map publishes a GeoJSON MultiLineString of the river reaches that lie downstream of any currently-active or last-48-hour CSO discharge in that operator’s network, refreshed every fifteen to twenty minutes. The geometry is the result; the topology graph traversal happens upstream of us. The map intersects each designated bathing water against these reaches and uses the answer in the status engine’s proximity check.

The dataset is published under Creative Commons Attribution-ShareAlike 4.0. The methodology page and the data-source footer credit Lipp, Dawe, and Balaji directly, and the project uses the data with their author’s permission. The underlying EDM data is the same Stream Storm Overflow Hub, Thames Water API and Scottish Water API the map already ingests independently — sewage-map is a derived layer, not a new primary source.

What this asymmetry means. A bathing water in the Southern Water service area has historical discharge events backing the map’s verdict. A bathing water in the Anglian service area has at most one event per overflow — the most recent one — and no historical context at all. Both produce honest verdicts under the map’s rules, but the depth of evidence behind those verdicts varies by orders of magnitude depending on which company operates the local infrastructure.

The map does not penalise spots in less-transparent operators’ footprints, but it does not pretend the data is equivalent. Where event history is shallow, the map’s verdict rests on the present-tense state of the overflow rather than on patterns observable over years. The water companies that publish more, the map can say more about.

Where licensing is explicit, and where it isn’t. Of the thirteen sources this map ingests, eleven publish their data under explicit open licences. The Environment Agency and Natural Resources Wales publish their bathing-water data under the Open Government Licence v3.0, the standard Crown-copyright open licence used across UK government data.

Seven of the English water and sewerage companies — Anglian Water Services, Northumbrian Water, Severn Trent Water, South West Water, United Utilities, Wessex Water, and Yorkshire Water — each publish their Storm Overflow Activity datasets under Creative Commons Attribution 4.0 (CC BY 4.0), declared in each operator’s own dataset metadata on ArcGIS Online. Wessex Water and Yorkshire Water additionally publish event-level history feeds under the same licence. Alex Lipp’s thames-sewage mirror is also CC BY 4.0, retained as a verification fallback during the project’s transition to Thames Water’s first-party API.

Two operators publish under bespoke open licences with the same shape as CC BY 4.0 but in the operator’s own words. Thames Water Utilities Limited publishes its first-party Open Data API under its own Open Data licence: a perpetual, royalty-free, worldwide grant that requires attribution, a link to the licence, and disclosure of any changes made to the data. Scottish Water publishes its near-real-time overflow API under its own open licence, with attribution required and Scottish Water cited as the data source.

Two sources — Welsh Water (Dŵr Cymru)’s Spill_Prod ArcGIS layer and Southern Water’s Rivers and Seas Watch service — do not currently carry an explicit licence declaration on the specific data layer this map ingests, although both operators publish company-level open-data commitments and Southern’s Annual Performance Report is explicitly CC BY 4.0. The methodology treats both as public regulator-mandated disclosure and credits the operator directly; explicit confirmation has been requested from each.

One regulator — the Scottish Environment Protection Agency — publishes via server-rendered HTML pages without a machine-readable licence declaration on the pages themselves, although SEPA data is Crown-copyright under the same default framework as the EA and NRW.

What attribution requires. Across the four explicit open-licence types this map relies on — OGL v3.0, CC BY 4.0, Thames Water’s bespoke Open Data licence, and Scottish Water’s bespoke open licence — the common requirement is attribution to the publisher. The map credits each operator in its detail panel where their data drives a verdict, and lists every source in the page footer with its licence. CC BY 4.0 additionally requires that changes made to the data are disclosed: this map pairs the operators’ published Start and Stop alerts into single events, assigns a 48-hour inferred closure to unmatched Starts, and collapses duplicate per-outfall rows by external ID. Those transformations are documented in the footer and apply uniformly across every operator’s data.

How status is decided

Every spot on the map has its status computed by the same rule set, evaluated in the same priority order. There are no special cases by region, no scoring algorithms, no machine learning. The rules read the regulator’s most recent classification, the date of the most recent sample, the most recent risk forecast, and the recent discharge activity from nearby storm overflows, and they reach a verdict.

The rules are evaluated in priority order. The first rule that matches produces the verdict; later rules are not consulted. The order matters because it encodes the project’s editorial position on what evidence outweighs what — an active sewage discharge nearby is stronger evidence than a favourable rolling classification, for example.

The eleven rules, in order.

If a sewage overflow whose discharge bears on a bathing water is currently discharging, the bathing water is flagged unsafe. The map asks three independent questions in priority order. First: does the operator’s own published linkage name this bathing water as affected by this overflow? Welsh Water, Wessex Water and Yorkshire Water publish such linkages for the overflows they consider operationally tied to specific bathing waters, and where present that assessment overrides the map’s spatial inference. Second: is the overflow within 5 kilometres Euclidean of the bathing water? Third: does the bathing water sit on a river reach downstream of an active discharge in the operator’s catchment, as published by sewage-map? Any one of the three is sufficient. The compound check is intentional: each captures a class of cases the others miss. As of right now, designated bathing waters meet the 5 kilometre condition specifically; further spots are flagged unsafe via the operator-linkage or catchment signals.

If the regulator’s annual classification for a bathing water is Poor, the bathing water is flagged unsafe regardless of recent sampling. A Poor classification is a four-year rolling statistic — the regulator is saying that, across the previous four bathing seasons combined, this water has failed water-quality tests often enough to fail at the Poor threshold. That assessment does not need a fresh sample to remain credible.

If the regulator has formally closed a bathing water, the map flags it unsafe and shows the closure as the reason. Only the Environment Agency uses this status.

If a sewage discharge has been recorded within the last 48 hours whose effect bears on the bathing water, the bathing water is flagged caution. Same three-tier proximity check as rule 1, applied to the recent-discharge window: operator-published linkage first, then 5 kilometres Euclidean, then catchment-aware downstream. The 48-hour window matches the de-facto industry standard used by Surfers Against Sewage and the Rivers Trust for “recent enough to still affect water quality.”

If the regulator has issued a short-term pollution risk forecast of “increased risk” and that forecast has not expired, the bathing water is flagged caution. The Environment Agency issues these forecasts daily during the bathing season for every classified English bathing water; Natural Resources Wales issues them for about of Welsh bathing waters; SEPA does not issue them at all.

If the regulator’s annual classification is Sufficient — the regulatory pass mark — the bathing water is flagged caution. Sufficient is the legally-acceptable lower bound: a water can pass at Sufficient and still spend much of the bathing season below the threshold any swimmer would consider “good.” Calling it caution rather than safe is a deliberate editorial choice. The classification is the regulator’s; the verdict is ours.

If the regulator’s classification is Excellent or Good, a fresh sample backs that classification (within 6 weeks), and the regulator’s monitoring season is currently active, the bathing water is flagged safe. This is the only path to a green pin. The freshness requirement is what makes the green pin meaningful — without it, every bathing water with a favourable rolling classification would render as safe year-round, regardless of whether anyone has actually tested the water in the last six months.

If the regulator’s classification is Excellent or Good but no fresh signal exists, the bathing water is flagged monitored — no current concerns. This is the middle state: the rolling assessment is favourable, but we will not call it safe today without current evidence. of Britain’s bathing waters sit here at this moment, awaiting fresh samples from the 2026 season.

If the bathing water is outside its regulator’s monitoring season, the map flags it unknown with the reason “outside the regulator’s monitoring season.” Only Scotland has a statutory monitoring season that ends — between mid-September and mid-May, every SEPA spot falls into this state.

If none of the above rules match, the map flags the bathing water unknown with the reason “insufficient data to determine swim safety.”

The thresholds. Three numeric thresholds appear in the rules above: 5 kilometres, 48 hours, 6 weeks. Each is documented as a constant in the code rather than buried in rule bodies, which means future revision is transparent.

5 kilometres is the geometric proximity radius — how close a storm overflow must be to a bathing water, on a straight-line map, for that overflow’s activity to bear on the bathing water’s status. The choice is geometric rather than hydrological; it captures coastal-overflow effects in calm conditions but misses river-catchment effects where a sewage discharge into a tributary thirty kilometres upstream of a freshwater bathing water affects it within the 48-hour window without ever coming within 5 kilometres. The catchment-aware signal from sewage-map closes that gap by checking whether the bathing water sits on a river reach already-downstream of an active or recent discharge. The two checks are evaluated together; either is sufficient to fire the unsafe (rule 1) or caution (rule 4) verdict.

48 hours is the recent-discharge window. It matches the standard used by Surfers Against Sewage and the Rivers Trust for the period during which a recorded discharge remains operationally relevant to water quality.

6 weeks is the freshness window. A regulator sample within 6 weeks counts as fresh enough to support a present-tense safety claim; a sample older than that is, for the purposes of this map, stale. Six weeks brackets typical EA in-season sample frequency plus latency through the publishing pipeline.

The confidence band. Each verdict the map produces also carries a confidence level — high, medium, or low. The level reflects how complete the underlying evidence is, not how strong the verdict is.

A verdict at high confidence rests on fresh, complete evidence: a recent regulator sample, a current classification, and discharge data from nearby overflows. Most caution and unsafe verdicts run at high confidence because the event triggering them is itself fresh evidence. Most safe verdicts today run at high confidence because the fresh samples backing them are recent (within four weeks).

A verdict at medium confidence has gaps — typically a sample 4 to 6 weeks old, or a risk forecast supporting a claim where a sample would normally also be expected.

A verdict at low confidence has substantial gaps — typically the most recent regulator sample is months old, or the spot is outside its monitoring season, or other key signals are missing.

The freshness requirement applies asymmetrically. Positive claims of safety require fresh evidence. Negative claims — a Poor classification driving an unsafe verdict, a Sufficient classification driving a caution — do not, because the multi-year rolling statistic is the evidence. The classification was earned over four years; it does not lose validity because no one happened to sample the water this week.

What we know we don’t know

The map has known limitations. The rest of this section catalogues them — not because the project is unsure of itself, but because a methodology page that pretends to have no limitations is one a reader cannot trust.

The proximity check is layered but not exhaustive. The status engine asks three questions in priority order: does the operator’s own published linkage name this bathing water as affected by this overflow (Welsh, Wessex, Yorkshire publish such linkages); is the overflow within 5 kilometres Euclidean; does the bathing water sit on a river reach published as downstream-impacted by sewage-map. Any of the three is sufficient. The operator-linkage layer is the most authoritative because the operators have already done the hydrological work; the 5-kilometre radius covers coastal cases adjacent to a CSO; the catchment signal covers river-network cases the radius would miss.

Limitations remain. The operator-linkage layer covers only three operators — the other eight publish no equivalent linkage data, so bathing waters in those service areas rely on the 5-kilometre and catchment checks alone. The catchment signal trusts sewage-map’s river-network topology to be complete; gaps in that topology (small tributaries it doesn’t encode, tidal connections across estuaries) propagate into our verdicts. And the radius fallback applies to bathing waters not near any modelled river reach, so a coastal beach far from a freshwater inflow continues to rely on the 5-kilometre check alone. Operator-published names don’t always match the regulator-canonical names — fourteen operator-published names currently fail to resolve to any spot in the map’s schema (the Bridlington Bay vs Beach mismatch, a handful of river-only operator references like “River Tawe” that aren’t designated bathing waters, and town-level Yorkshire references like “Ilkley” that map to a more specific regulator name like “Wharfe at Cromwheel, Ilkley”).

Only one water company in Britain publishes per-event review classifications. Southern Water reviews its own storm overflow events and publishes a flag for each one indicating whether the company considers it a genuine sewage discharge or a probable monitor false-trip. About of Southern’s recorded events carry the “Not Genuine” flag, meaning a problem with the monitoring sensor rather than an actual spill. The map respects this distinction — Southern’s Not Genuine events do not contribute to caution or unsafe verdicts.

None of the other ten UK water and sewerage companies publishes a comparable signal. Thames Water, the Stream-format snapshot operators, Welsh Water, Scottish Water — all publish their EDM events without indicating which they themselves consider genuine versus false-positive. The map applies no equivalent filter to events from those operators, because there is no equivalent data to apply. This means a bathing water in any operator other than Southern’s footprint may show caution from a discharge event that the operator would, if asked, describe as a sensor anomaly. The map errs in the direction of overstating risk rather than silently applying Southern’s quality control to data the other operators have not classified. The asymmetry is a property of what operators choose to publish, not of the map’s reading of that data.

The regulators publish at unequal quality. The Environment Agency, Natural Resources Wales, and the Scottish Environment Protection Agency implement the same European Bathing Water Directive heritage but produce structurally different qualities of public data. The EA publishes a documented API with classifications for every designated water and short-term risk forecasts for every classified water during the season. NRW uses the same Defra infrastructure but publishes risk forecasts for only of Welsh bathing waters. SEPA publishes via server-rendered HTML rather than an API, monitors only between 15 May and 15 September, and does not publish risk forecasts at all. One of SEPA’s designated bathing waters carries a published grid reference that cannot be parsed — Lunderston Bay, with a 9-digit grid reference that violates standard Ordnance Survey conventions. Until SEPA corrects its own publication, that location cannot appear on this map.

The map does not penalise spots in lower-publishing regulators’ footprints, but it does not pretend the data is equivalent.

Some operators publish years of event history; others publish a single most-recent event. Southern Water has published five and a half years of event-level history. Thames Water publishes about four years. Wessex Water about sixteen months. Yorkshire Water about four months. Five major English water companies — Anglian, Northumbrian, Severn Trent, South West, and United Utilities — publish only the current state of each overflow plus its single most recent event, with no historical depth available. Welsh Water publishes the latest event per overflow. Scottish Water publishes the current and previous event per overflow. The depth of evidence varies by orders of magnitude depending on which company operates the local infrastructure: a bathing water in the Southern Water service area is backed by hundreds of thousands of historical events; a bathing water in the Anglian service area is backed by at most one event per nearby overflow. Both produce honest verdicts under the map’s rules, but the map does not pretend the data is equivalent.

The map does not display this asymmetry on the pin colour, but it shapes what the map can and cannot say about historical patterns at any given spot. Where event history is shallow, the map’s verdict rests on the present-tense state of the overflow. Where it runs deep, the verdict can be situated in a longer record.

The map covers only designated bathing waters. The spots on the map are the swimming locations that have been officially designated by their regulator. The Outdoor Swimming Society, Wild Swimming guides, and OpenStreetMap collectively list thousands more places where Britons actually swim — wild swimming spots on rivers and lakes, urban lidos, swimming holes on rivers Britons have swum in for generations. None of these are on the map, because none of them have a regulator-published classification to anchor a status decision.

The River Thames is a useful illustration of where the boundary falls. Pangbourne Meadow and Ham & Kingston have just been designated for 2026 and appear on the map without classifications yet. The stretch through Henley, by contrast, is not a designated bathing water; Henley swimmers exist in numbers but there is no regulator-published water quality data for that stretch. The absence of Henley from the map should not be read as a statement about whether it is safe to swim there. It should be read as a statement about what the regulators have chosen to monitor.

A future phase of the project would extend coverage to non-designated spots, with appropriately different editorial framing — a non-designated spot has no rolling classification, no risk forecast, and the map’s status logic would need to adapt to the different evidentiary base.

The freshness window itself is editorial. The map treats regulator samples up to six weeks old as fresh enough to support a safety verdict. Past six weeks, a sample no longer counts. The threshold is documented in code as a named constant rather than buried in rule logic, so future revision is transparent. But six weeks is itself an editorial choice — long enough to capture typical EA in-season sample frequency and publishing latency, short enough to ensure a present-tense safety claim is grounded in present-tense evidence. Reasonable readers could disagree about where the line should sit. The map’s position is that it should be visible.

Things the project has found

Building this map surfaced a handful of observations about how British water-quality monitoring actually works. They are not the methodology of the project; they are the consequences of looking carefully at what the regulators and water companies publish.

Sampling cadence is uneven across the year. The statutory English bathing-water season begins on 15 May; Natural Resources Wales has typically begun publishing its 2026 samples a week or two before the EA does. In any given spring, the result is that Welsh spots return to confident-safe verdicts before English ones, and by midsummer the asymmetry narrows. By mid-September to mid-October, samples taper off again. This is not a data-ingest delay; it is the structural rhythm of how the bathing-water monitoring season works, and the map surfaces that rhythm rather than hiding it.

Every English bathing water enters May with samples from the previous autumn. Between the end of one English bathing-water season (late September to mid-November) and the first samples of the next (mid- to late May), every English bathing water’s most recent regulator sample is months old. This is not a data-ingest delay; it is the structural reality of how the bathing-water monitoring season works. Until 2026 in-season samples have begun landing, every map claim about an English bathing water rests on samples from the previous autumn, and the methodology surfaces that staleness rather than masking it.

Scotland’s monitoring season ends. The Scottish Environment Protection Agency samples bathing waters only between 15 May and 15 September each year. For the other eight months, SEPA does not sample, does not publish classifications updates, does not issue risk forecasts. Between mid-September and mid-May the following year, every Scottish bathing water on the map renders as “unknown — outside the regulator’s monitoring season.” Two-thirds of the calendar year, the map cannot say anything specific about Scottish water quality because the regulator is not measuring it.

Operator transparency varies enormously. Of the eleven UK water and sewerage companies, exactly one — Southern Water — publishes a per-event flag indicating which of its recorded sewage events the company considers genuine versus probable monitor false-trip. About of Southern’s recorded events carry the “Not Genuine” flag. The other ten companies publish their event data without any equivalent classification. This is a structural finding about UK water-company data transparency: the asymmetry is not in the map’s reading of the data but in what operators choose to publish. The single operator that does it shows it can be done.

Welsh Water publishes overflow-to-bathing-water linkages that the map’s geometric model misses. Welsh Water — the only operator in Britain to do so — publishes, for each of its overflows, a curated list of which bathing waters that overflow is operationally linked to. The map’s 5-kilometre geometric proximity model agrees with about 89% of Welsh Water’s published linkages — 225 of 253 overflow-to-bathing-water pairs fall within five kilometres. Two of the remaining linkages fall beyond the map’s 5-kilometre radius and are visible only in Welsh Water’s data, not in the map’s geometric inference. These two — Penarth Kymin to Jackson’s Bay at 8.6 kilometres, Rhoose to Whitmore Bay at 5.5 kilometres — are direct evidence from the operators themselves that catchment effects beyond simple Euclidean distance matter. The map’s future direction will close this gap.

Thames Water flags monitor-offline events distinctly from monitored discharges. Thames Water’s data feed distinguishes between “ Start” and “Offline Start” alerts — the latter indicating that the company’s Event Duration Monitoring infrastructure itself was offline during a recorded event. Of the archived Thames events ingested so far, are Offline Start alerts rather than monitored discharges. This is itself a transparency choice — Thames Water could have published all its events as ordinary Start alerts and the methodology would have no way to distinguish monitor-offline events from monitor-online ones. Publishing the distinction is editorially significant; the methodology preserves it.

SEPA’s own publication contains an unparseable grid reference. Lunderston Bay, one of Scotland’s designated bathing waters, has a published OS grid reference of nine digits — NS204074514. OS grid references must contain an even number of digits. This particular bathing water cannot be plotted on any map by any consumer of SEPA’s published data. The fix lives entirely on SEPA’s side. The map excludes Lunderston Bay rather than guess at a corrected coordinate, and the methodology surfaces the exclusion rather than silently dropping it.

Most currently-active sewage discharges are not near designated bathing waters. At this moment, storm overflows are actively discharging sewage across Britain — a count that varies hour by hour, much higher after rainfall, lower in dry weather. The vast majority are inland sewage works, urban combined sewer overflows distant from coast, and treatment-works overflows in towns that are not designated bathing-water locations. Of those discharges, bathing waters are currently within five kilometres of one. The map flags those few cases — that is what designated bathing waters are for — but the picture of “currently discharging sewage” is much larger than what the map’s pin states reveal. A future iteration of the map would surface the wider discharge activity directly.

The depth of historical event data varies across operators by orders of magnitude. Wessex Water has published historical events; Southern Water has published . At the other end, the seven English water companies that publish only the latest event per overflow expose at most one event per outfall. A bathing water in the Wessex service area can be situated against hundreds of historical events; a bathing water in the Anglian service area sits against one. The map does not penalise this asymmetry, but it does not pretend the data is equivalent. The methodology page is the place this difference is named.

The full architecture and deferred enhancements for the project — features, data sources, and audit observations that are not yet in production — are documented in the project’s public backlog. For corrections, questions, or partnership conversations, contact hello@swimmable.co.uk.

Methodology last refreshed: 5 Jun 2026, 02:16