A água que se perde, e a que nunca foi medida

Corrected 29 July 2026The industry figure on this page was wrong, and is fixed. It previously read 609 hm³ / 10.7%, which counted 300 hm³ of seawater and estuarine abstraction — coastal aquaculture and the Sines industrial complex — as freshwater use. APA excludes seawater from its own published chart and states the mechanism verbatim for energy; applying the same test to industry closes the gap to 1.2 hm³. Industry is 309 hm³ / 5.7% of freshwater consumptive abstraction. Agriculture correspondingly rises from 73.7% to 77.8%. This is the same class of error the page warns about in the hydro figures — caught there, missed here.

Before asking what leaks, establish what there is. The answer depends entirely on whether you count the water that passes through turbines — and that single choice changes every percentage by a factor of twenty-one.

Portugal abstracted 121,968 hm³ of water in 2018 on the gross measure. It abstracted 5,417 hm³ of fresh water for uses that actually consume water. Both numbers are correct. The difference is hydroelectricity, which is 94% of the gross figure and is returned to the river in full — and which is counted more than once, because the Douro cascade records the same water at each successive dam. Douro basin "abstraction" exceeds the river's entire annual flow.

Use the toggle to see what that does to every sector's share.

Press onlyThe figure everyone quotes does not exist. The set in circulation — agriculture 75%, urban 13%, industry 5%, tourism 0.7% — appears in no single source and sums to 93.7%. It splices PNA 2015's agriculture figure onto a different document's urban figure with a different denominator. The real PNA 2015 set is 75 / 20 / 5 / 0.7.

The 75% itself survives scrutiny. It traces to Plano Nacional da Água 2015, Quadro 7, describing water use in 2012 excluding energy. The independent 2018 reconstruction above puts agriculture at 77.8% of fresh water — close corroboration from separate data six years later.

Not measuredAgriculture's number is a model, not a measurement. The PGRH computes it as irrigated area × crop coefficient ÷ conveyance efficiency. 71% of Portuguese farmers have no water meter — a figure from a Gulbenkian/Ipsos survey of about 490 holdings (2021), not from INE's Recenseamento Agrícola, which does not contain this statistic (attribution corrected 29 July 2026). The sectors that are metered — urban and industry, from tax declarations — agree across sources to within 2.5% where the comparison was possible. Agriculture does not: in the Guadiana, the volume declared for the water-resources levy is 62% below the figure modelled in the PGRH, and between planning cycles the regional differences reach +98%.

This is the part that is genuinely, carefully measured — 214 retail operators in the regulator's universe — 14 of which did not report the indicator in 2024 — published annually. It covers about 14% of the water Portugal consumes.

Água não faturada — non-revenue water — is not a synonym for leakage, and ERSAR is explicit about it. The indicator is the gap between water entering the system and water billed, and it contains three different things.

So of every hundred litres entering a Portuguese distribution network, about twenty leak into the ground, roughly four are lost to metering error and unauthorised use, and about three are delivered deliberately but never billed — firefighting, street washing, municipal irrigation.

Watch the denominatorDo not average the column. The unweighted mean across operators is 36.6%. The volume-weighted national figure is 26.4%. Small operators with terrible ratios are numerous but move little water. Averaging the operator column overstates national loss by ten percentage points.

The variation between operators is the real story, and it is enormous — from 2.9% to 72.7%. Two neighbouring municipalities can differ by fifty points.

Corrected 29 July 2026The figures in this note were wrong and have been corrected. It read 15.3% against 28.5%, a pair that does not reproduce from the per-entity workbook under any aggregation. Operators under concession show 14.9% non-revenue water against 28.0% for the rest, weighting by volume — on a simple average, 18.8% against 39.0%. ERSAR publishes no such comparison — it was computed from the per-operator rows, and it carries a confounder: concessions do not serve a random sample of Portugal. They cluster in denser, newer, more coastal networks. The gap is real; the causal reading is not available from this data.

The question has an official answer in Portugal, and the answer is that nobody has counted.

PENSAARP 2030 — the national strategy in force — defines precisely the right indicator: "percentagem de rede que excedeu a vida útil esperada", the share of network past its expected useful life, with a target of under 10%. Its current value is recorded as s/d: sem dados. No data. The same for wastewater, and for stormwater, whose network the plan notes has never been inventoried at all.

Not measuredThe indicator that would answer this question has existed on paper since 2024, with a target attached. The measurement behind it does not exist. It was already the diagnosis in the predecessor plan a decade earlier.

Two things are published, and together they support an answer.

First, ERSAR records for each operator the length of mains more than ten years old — it is the hidden denominator of the rehabilitation indicator. Nationally that is 95,924 km of 110,327 km: 86.9%. Ten years is a low bar against a forty-year design life, so this says the network is not new; it cannot say how old it is.

Second, the rate of replacement. And this is where the picture becomes clear.

Inference — arithmetic shown 2,304 km rehabilitated over five years ÷ 5 ÷ 110,327 km = 0.418% per year.
At that rate a full replacement cycle takes 239 years. PENSAAR 2020 assumes a 40-year design life for water mains. Portugal is therefore replacing its network roughly six times slower than its own assumed design life requires. The plan's own words, a decade ago, were that renewal ran at "praticamente a metade do que as boas práticas recomendam" — practically half what good practice recommends. It has since halved again.

This is the honest answer to the age question. Not "the average pipe is X years old" — that number does not exist — but a network being replaced on a two-century cycle against a forty-year design life, in which 87% of mains have already passed ten years, and whose infrastructure value index has fallen from 0.37 to 0.36, below ERSAR's acceptable floor of 0.40 and still declining.

Portugal has costed this. The number is in the Diário da República.

PENSAARP 2030 requires €5,500 million to 2030 across water supply, wastewater and stormwater, of which €2,860 million — 51.8% — is rehabilitation. Of the total, €1,832m falls on state-owned systems and roughly €3,565m on municipal ones. The entire swing between the plan's ambition scenarios sits in municipal rehabilitation.

Inference — arithmetic shown What full replacement would cost. ERSAR reports 110,327 km of distribution mains plus 10,665 km of bulk mains — 120,992 km. PENSAAR 2020 gives a unit renewal cost of €80 per metre. Multiplying: €9.68 billion to replace the water network once.
Two caveats that matter. The €80/m is in roughly 2014 prices and understates 2026 costs. And this is full replacement — the share actually needing replacement is exactly the figure recorded as s/d in section 03.

Against a stated requirement of €5.5bn, the PRR's dedicated water component — Componente 9, Gestão Hídrica — has €239.9m approved and €126.8m paid (snapshot of 29 July 2026). And it reaches 25 of Portugal's 308 municipalities.

Almost all of it lands in two places: the Algarve (€104.0m across 16 concelhos, 71.6% paid) and Madeira (€82.2m across 7, 23.8% paid). One Alentejo municipality, Mértola, and one Lisbon municipality, Oeiras, account for the rest. 283 municipalities receive nothing from this component.

Read carefullyThis is not the whole picture, and it would be wrong to say Portugal funds only 4% of its water need. PENSAARP is financed mainly through tariffs, municipal budgets and the Portugal 2030 regional programmes; the PRR component is one targeted instrument aimed at water-scarce regions, not the sector's budget. What the comparison does show is how small and how geographically concentrated the headline-grabbing money is.

Checked against the dataGovernment communications state that Algarve water operators "executaram mais de 40 milhões de euros de investimento na reabilitação de redes". That claim maps to one specific sub-line — C09-i01.01, the AMAL-managed municipal loss-reduction programme, 51 projects across 15 of the 16 Algarve concelhos — and the verdict splits. Approved: €47.65m, which confirms it. Paid: €35.97m, which refutes it by €4.03m. The claim's own verb is executaram, and because this sub-line is financed 100% by Next Generation EU with no municipal co-funding, executed investment and disbursed grant are the same euros. On the measure the wording implies, the register does not support €40m. The approvals figure also already exceeds AMAL's own stated envelope of €43.9m.

UnverifiableThe companion claim — 160 km of network intervened against a 125 km target — cannot be tested from the register. The target exists as milestone 9.2, "pelo menos 125 km de intervenções destinadas a reduzir as perdas de água no setor urbano no Algarve", due Q1 2026. In the file that is supposed to record it, both the achieved value and the completion date are empty, and no structured kilometre field exists anywhere in the eleven PRR datasets.

Nothing recorded as delivered C09 has concluded 0 of its 7 milestones — the only one of the PRR's 22 components at zero. 76 of its 77 projects carry no actual completion date at all, against 30.7% nationally, and 65 are already past their planned date, carrying €156.2m approved. The one project with a recorded completion finished 517 days late and is worth €14,466.

Not all of it is water€35.0m of the €239.9m is a photovoltaic park at Alqueva, sitting inside the water component. Reading all of C09 as water money overstates it by 17.1% (the park is 14.6% of the total). And the "Plano Regional de Eficiência Hídrica do Algarve" spends its largest single location share — €18.6m, 36.3% — in Mértola, which is in the Alentejo, not the Algarve.

No consequencePENSAARP 2030 quotes the World Bank's verdict on its own predecessor: "Inconsequência dos alertas para os desvios entre resultados e metas" — the alerts about deviations between results and targets had no consequences. Searching the new plan for consequência, incumprimento, sanção and penalização finds no consequence mechanism either. The monitoring report that would show progress was never published — the page the RCM designates is up, but carries only the plan's volumes and the RCMs themselves — as is the annual public report the plan's own oversight group committed to publishing.

InferenceFrom 28.8% to 26.4% is 2.4 points of the 10.8 needed — 22% of the distance in the first two of the eight years of data (2022→2024): a pace of 1.2 points a year against the 1.35 the target requires. Meeting the 2030 target means stopping the loss of a further 71 hm³ every year: about 8.4% of everything the country's public networks currently put into supply.

The comparison people reach for cannot be made. That is worth stating plainly, because the sentence "Portugal is among the worst in Europe for water loss" is common and is not currently supported by any complete table.

On scarcity Portugal is measurably stressed: 7.27% against an EU27 aggregate of 5.15%, some 40% above. But the national annual figure conceals the thing that matters. APA's own basin-level assessment puts Sado e Mira at 74%, Ribeiras do Algarve at 66% and Guadiana at 51% — all in severe or extreme scarcity — with summer values reaching 84%.

Portugal is absentOn network losses, Portugal is not in the comparison at all. EurEau's Europe's Water in Figures 2026, the only EU-wide table of network losses that exists, lists Portugal as N/A for both real losses and non-revenue water. Eurostat has a leakage category in env_wat_bal; twenty countries report into it and Portugal reports none. Eurostat's Portuguese abstraction data stops in 2020 despite metadata advertising 2023.

There is a mechanism that would fix this. The recast Drinking Water Directive, Article 4(3), required every member state to assess leakage on a common methodology and report to the Commission by 12 January 2026. That deadline passed six months before this was written. Probing the Commission's own actions tracker, CIRCABC and WISE found no published output.

UncitedThe European Commission's own Water Resilience Strategy states that "current national leakage levels vary from 8% to 57%". The claim carries no footnote — the footnotes either side concern unrelated matters. It reads as an authoritative EU-wide fact and cannot be traced to country data from the document itself, so it is not used here to place Portugal anywhere.

The deeper problem is that three incompatible definitions are in play at once: technical real losses, non-revenue water, and Portugal's água não faturada, which is an economic indicator rather than the leakage index the Directive requires. The 2021 edition of EurEau's report says so verbatim — "it is not reliable to make comparisons between countries in this context" — and the 2026 edition keeps a more general warning.

Agriculture is three-quarters of Portugal's consumptive water. It is also the least measured.

DGADR publishes real conveyance efficiency for the 33 largest public irrigation schemes: a simple average of 75–76% in 2024, ranging from 8% at Alvor to 98% at Alfândega da Fé, with Alqueva at 92%. That is genuine per-scheme government data and it is better than nothing.

Corrected 29 July 2026This note gave ~95% as the volume-weighted efficiency; the ratio is not valid. It comes from dividing totals with unequal coverage — about 120 hm³ of consumption belongs to schemes with no reported abstraction. Comparing only the schemes that carry both figures, the volume-weighted efficiency is 83–85%, alongside the 75–76% simple average. Alqueva alone is about 40% of national volume. Quoting the simple average on its own still misrepresents the system.

The asymmetryIn APA's Guadiana basin management plan, urban losses have a regulator, a year and a figure: physical losses 25.3% regionally, 21.2% nationally, 2018, sourced to ERSAR (Part 3). In the objectives table of the same plan (Part 5), the agricultural row has no regulator, no baseline and no figure — only a 2027 target of 25%. Portugal's own planning document quantifies the losses of one-eighth of its water and leaves the three-quarters blank.

The circularity trap is worth naming, because it is easy to fall into. The obvious move is to multiply agriculture's 4,212 hm³ by DGADR's 75% efficiency to get irrigation losses. That would be wrong: the PGRH abstraction figure is already derived by dividing field consumption by conveyance efficiency. Applying the efficiency twice manufactures a large, confident, entirely invented number. No such figure appears on this page.

Two findings from that table deserve pulling out. Data-centre cooling at Sines is seawater and non-consumptive — confirmed from APA's own environmental filings, not from press. But the enforceable licence cap today is 17.52 hm³/yr while the full-campus design maximum extrapolates to around 531 hm³/yr: a factor of thirty between what is permitted now and what is designed for. And under the EU Energy Efficiency Directive's data-centre register, Portugal reported 4 of roughly 42 qualifying facilities — about 10% compliance — one of ten member states at or below that level.

Meanwhile water reuse stands at 8.3 hm³, or 1.1% of treated wastewater, against a government target of 116 hm³ by 2040 — a fourteen-fold increase required in fourteen years.

What follows are the things this page cannot tell you, and why.

Average pipe age does not exist. No Portuguese document states it. The 239-year replacement cycle is an inference from rehabilitation volumes, not an asset survey.

The map is by operator, not by municipality. ERSAR reports on entidades gestoras. Where one serves several concelhos, each carries the same figure. 59 of 308 municipalities have no figure at all — 30 of them because ERSAR's mainland list does not cover the Azores or Madeira. These are drawn hatched.

The public–private gap is a calculation, not an ERSAR finding, and confounded by where concessions operate.

Irrigation efficiency is self-reported by scheme managers, with no independent regulator, and DGADR's own report calls for the measurement campaigns and telemetry that would make it comparable to ERSAR's data.

Sector shares are 2018. The PGRH third cycle is the most recent full reconstruction available; the fourth cycle is not published.

The PRR figures have since been independently re-derived, and the exercise found an error in this page. A second pass rebuilt them from the source registers by a different route, reconciled the attribution rule across all 77 projects with zero mismatches, and reproduced the per-concelho file byte for byte. It also compared two snapshots three days apart, which agreed to the cent. What it corrected was section 04's reading of the €40m claim: this page originally set that claim against the whole Algarve C09 total and concluded the press figure was conservative. It maps to one sub-line, and on the payment measure the register does not support it. The corrected verdict now stands in section 04.

Full audit of 29 July 2026. The 69 facts on this page were checked against the primary documents — RASARP 2025 and Anexo X, PGRH, RCM 23/2024, the PRR files, the Eurostat API, EurEau, DGADR, INE — by seven independent agents, with adversarial counter-proof of every negative verdict. 53 verified, most of them to the decimal, including every claim of absence, tested with positive-control probes. The resulting corrections are marked where they occur: the concession ratios (section 02), the attribution of the water-meter figure and the WEI+ threshold (section 01), the −35%/+75% range, replaced by checkable examples, the progress passage and the framing of the monitoring report (section 04), and the weighted irrigation efficiency (section 06). The PRR paid total was updated to the same day's snapshot.

On methodEvery figure on this page carries a source, a year and a type. Where a source could not be found, the probes attempted are recorded rather than the gap being filled. Nothing here is drawn from general knowledge, and no number appears without a document behind it. Where a figure is derived, the arithmetic is shown so it can be checked and overturned.