Finding leaks from orbit: how network loss became a data problem

Utilities worldwide lose about 346 million cubic metres of non-revenue water every day, worth a conservative US$39 billion a year.¹ Several countries in this region report losing around half their supply: Jordan’s kingdom-wide non-revenue water was 50% in 2022.² What has changed is the instruments for finding it.

Three layers of seeing

The classic instrument is the district metered area, a discrete part of the network usually created by closing valves; in UK urban areas, the IWA guidance notes, such districts vary between 500 and 3,000 properties. Meter the feeds into the district and read the minimum night flow over a one-hour period, when customer consumption is usually low.⁴ Whatever still moves, after an allowance for legitimate night use, is mostly leaking. The method gives a district a loss figure. It does not give a location.

Acoustics answer the where. A pressurised leak makes a noise that travels along the pipe: on US EPA figures, a leak of 5 gallons a minute can be picked up about 300 to 370 metres along a 6-inch iron main, and 120 to 180 metres along PVC of the same size. Paired sensors, or correlators, time the same sound arriving at two points to estimate where the leak lies, typically to better than a metre per 100 metres of pipe (US EPA).⁵ That accuracy is the difference between digging one hole in the road and twenty.

The newest layer is orbital. L-band synthetic-aperture radar can penetrate up to about two metres of soil, and ASTERRA says its algorithm identifies treated drinking-water leaks by their dielectric signature.⁶ Commercial services now sell city-wide leak surveys from satellite passes. ASTERRA says its surveys can detect leaks as small as half a litre a minute and reduce the area for field inspection to 5 to 10% of the pipe length; these are supplier claims.^3 6

The percentage trap

The headline number itself needs handling with care. A World Bank paper co-authored by Roland Liemberger constructs the example: a network supplied 12 hours a day reads 20% loss. Supply the identical network continuously, with not one leak repaired, and the reading rises to 29%, because water runs out of the same holes for twice as long.⁷ A utility can therefore improve its network and watch its headline number worsen.

Amman is the standing example. Miyahuna reported non-revenue water of 44.7% in 2023, but its network has supply only 22.6% of the time, and the volumetric series tells the opposite story: water loss per connection fell from 1.25 to 0.99 cubic metres a day between 2019 and 2023.⁸ The 2023 headline rise was an accounting event, the merger of Balqa governorate and its 62% losses into Miyahuna’s books. Litres per connection per day, read with supply time and pressure as the World Bank paper advises, is the measure that survives comparison.⁷ The percentage is the measure that makes headlines.

The meter is the other half

In some Jordanian governorates the apparent-loss component is the major contributor to high non-revenue water. Ageing mechanical meters under-register, connections go unbilled, bills go uncollected. Under intermittent supply the metering problem compounds: the EU’s Jordan policy review notes meters begin under-registering within two years in these conditions.² In March 2018 Amman’s utility, Miyahuna, signed a three-year contract for analytical services and 76,000 ultrasonic meters as part of its non-revenue-water reduction plan. The static meters have no moving parts and provide detailed data for detecting leaks or unusual usage patterns.⁹

What good looks like, with a caveat

Dubai is the standing rebuttal to the argument that the region’s heat and pressures make losses inevitable. DEWA reports water network losses of 4.5% for 2024 and 4.4% for 2025, which it calls the world’s lowest,¹⁰ and credits “various advanced technologies and systems”, among them more than 1.1 million smart water meters, for the fall from 42.5% in 1989.¹¹ The figure deserves one caveat. DEWA publishes no full water balance behind it, so the headline cannot be split into physical and commercial losses or tested against a volumetric measure.

On a volumetric measure, the gap between the two cities is wider than the headline percentages suggest: Miyahuna reported about 990 litres of water loss per connection per day in 2023,⁸ while our estimate from DEWA’s 2024 figures gives about 77 litres of total network loss per water customer per day in Dubai.^12 10 The commercial case follows from that arithmetic. Reducing leakage makes treated water available to customers and can be more cost-effective than developing new supplies.⁷

Why the money now flows through contracts, not grants

Donor finance came first, with mixed results. More than JOD 1.28 billion has been committed to non-revenue water in Jordan alone, according to an EU-funded review published in May 2024. It records a KfW programme of EUR 26.5 million in Karak with no impact on non-revenue water, with a post-evaluation reporting an increase; a KfW programme of EUR 40.2 million in the northern governorates that was not measured and had little impact; and a USAID programme of USD 60 million (FARA Phase I) whose gains were not sustained because operations staff breached the zonal boundaries.² The transferable lesson, on our reading, is that district metering depends on operators respecting the districts.

The procurement model has since shifted. Amman’s planned 24-year performance-based NRW project, developed with IFC support, is intended to engage a private-sector operator to reduce water losses and improve network efficiency; it remains in procurement.^13 14 15 Paying for outcomes rather than equipment is, on our reading, the design answer to programmes whose gains were not sustained.

In the region

Several utility programmes in the region carry a technology line. Jordan has two in the register: the Amman performance contract and the North Jordan NRW reduction programme. Saudi Arabia’s National Water Company runs distribution under management contracts with loss targets: its 2021 Eastern cluster contract counts reducing water loss among 14 key indicators.¹⁶ Egypt’s holding company runs a platform tracking its district-metered zones and meters,¹⁷ while Oman’s Nama is installing more than 400,000 smart meters.¹⁸ Some recent contracts combine metering or other hardware with digital monitoring, and Oman’s 2026 performance-based contract links operator remuneration to 33 key performance indicators, including cutting water losses from 34% to 11% by 2040.¹⁹

The case study

One case in this sector, examined in full.

Who is doing this

The buyers and operators behind the procurements this brief describes.

• MiyahunaUtility · Jordan
• Dubai Electricity & Water Authority (DEWA)Utility · United Arab Emirates
• National Water Company (NWC)Utility · Saudi Arabia
• Holding Company for Water & Wastewater (HCWW)Utility · Egypt

Mandates, procurement routes and the rest of the region’s institutions are in the decision-maker directory.

What to watch

• 2026Amman’s 24-year performance-based NRW project: the RFQ deadline was extended to 12 May 2026,²⁰ and three applicants were prequalified on 9 July 2026. The PPP Unit said the announcement did not constitute a contract award.¹⁵ US commercial reporting expects the request for proposals and the award within 2026.¹⁴ Progress is tracked on the procurement page.

Further reading

• Liemberger, R. & Wyatt, A. (2019), “Quantifying the global non-revenue water problem”, Water Supply 19(3). The global loss estimate this brief opens with. doi.org
• Kingdom, B., Liemberger, R. & Marin, P. (2006), The Challenge of Reducing Non-Revenue Water in Developing Countries, World Bank. The paper behind the percentage trap. documents1.worldbank.org
• LDK Consultants / EU WES, NRW Policy for Jordan (May 2024). Governorate tables and a review of what donor programmes achieved. wes-med.eu
• ASTERRA, Using Synthetic Aperture Radar for Underground Leak Detection (supplier white paper). asterra.io
• IFC press release (6 June 2024) on the agreement behind Amman’s performance-based NRW contract. ifc.org