Liwa: the reserve Abu Dhabi banked in the sand

Desalination has one structural weakness: it makes water just in time. Before this scheme, Abu Dhabi City held two to three days of drinking water.⁸ The Gulf is a shallow sea with weak exchange through the Strait of Hormuz,⁹ and its states produce about 40 per cent of the world’s desalinated water.¹⁰ Surface storage in that climate suffers high evaporation losses.⁵

An aquifer solves the evaporation problem and the land problem at once.^5 11 The Environment Agency in Abu Dhabi began investigating the idea in 2002.¹ The question was never whether water could be put underground. It was whether the same water could be got back out fit to drink.

Desalinated seawater from the Shuweihat plant is piped to a site in the Liwa desert, about 160 km from the coast.^1 7 It is infiltrated by gravity through three basins, each 50 metres across, into a dune-sand aquifer in the Al Dhafra region.^1 3 Recovery runs through 315 wells in three fields, reaching up to 80 metres below ground, with 155 km of pipework to the reservoirs.^3 1 5

The pilot changed the design, which is the part worth copying. Dual-purpose wells, using the same boreholes to inject and recover, showed signs of clogging. The basin did not.^12 3 Through 250 days of constant recharge at 250 cubic metres an hour there was no sign of declining capacity, and no drop in well performance.⁷

Dornier Consulting designed it with GIZ,⁸ an ACC-POSCO joint venture was among the contractors,¹ and Dar did the groundwater modelling.¹¹ TRANSCO, the utility authority’s transmission subsidiary, managed construction, and the finished reserve was unveiled in January 2018.^1 2 The Environment Agency funded the peer-reviewed water-quality modelling.³

The chemistry, which was the real risk. The native groundwater carries elevated salinity, chloride, sulphate, fluoride, sodium and chromate, and stored water in such schemes generally changes quality through redox reactions. Late in design it was feared a costly post-treatment plant would be needed.³

Monitoring found practically no redox reactions, little carbonate dissolution and low overall aquifer reactivity, in an aquifer that was already (sub)oxic.³ Pilot observations and modelling indicated that recovered water would meet Abu Dhabi’s drinking standards without the extra plant.^3 7

The stored water also moved little during the pilot. The outer edge of the freshwater body moved 0.2 metres a day across 250 days of recharge.¹² That is the parameter deciding whether a reserve is still there when it is wanted. Modelling then fed back into operations. The injection rate was raised by 20 per cent, and a continuous trickle of 568 cubic metres a day is to be infiltrated during storage to hold the bubble against regional drift.³

The efficiency figures in circulation have not been reported for the full-scale plant. The 85 per cent that circulates was a model output at design stage, stated as such by Dornier’s project hydrogeologist in 2012.⁸ The 85 to 90 per cent physically recovered came from the pilot, running at 250 cubic metres an hour, not from the 315-well plant.^12 7

The peer-reviewed modelling is candid about what it did not test. The full-scale recovery rate is six times the infiltration rate, which the authors warn could draw brackish or saline water up from the aquifer beneath. They list upconing, aquifer clogging by clay mobilisation and basin clogging as unassessed. They also note weak signs that clay mobilisation may already have occurred during the pilot.³

Quality also degrades with storage time. Their prediction is that after ten years the efficiency falls from about 85 to 60 per cent. Chromium reaches 70.7 micrograms a litre against a 50 microgram standard, with sodium, arsenic and fluoride also exceeding.³ A reserve held for a decade is not the same asset as a reserve held for a year.

The scale claim needs qualifying too. The launch publicity gave a fallback pumping capacity of 100 million gallons a day.^1 2 TRANSCO’s own planning statement gives a recovery of 40 million gallons a day,⁴ roughly 4 per cent of the 980 million the emirate used daily at launch.⁶ That is a genuine emergency reserve for one city, not a substitute supply.

What is missing from the record matters. TRANSCO’s planning statement describes recovery for emergency supply,⁴ but the sources reviewed report no full-scale recovery test, no published activation criteria and no independent audit or project-specific environmental assessment. Nor does any source reviewed report that the reserve was drawn on during the war that began in February 2026, when plants were struck elsewhere in the Gulf.^13 14 The cost is reported several ways: US$350 million, reported by Mohamed Dawoud of the Environment Agency Abu Dhabi,⁷ a construction value of AED 1.3 billion from one contractor,⁵ and an announced AED 1.61 billion, about US$435 million.^2 1

The transferability test is restrictive, and the evaluation literature says so. Writing for IWMI, the Environment Agency’s Mohamed Dawoud calls Liwa a benchmark case, then sets the conditions. In his words, the technology “holds promise particularly for Middle Eastern countries with strong economies, and strong linkages between water and energy production”.⁷

Two things have to be true. There has to be a structural surplus of desalinated water, which Abu Dhabi has from plants tied to power generation. And there has to be a balance sheet that can absorb roughly US$400 million for insurance that may never be claimed.^7 3

The geology is not a formality either. The chemical success rests on a low-reactivity, (sub)oxic, calcareous sand aquifer with a pH around 8.³ Import the engineering without the geology and the chromium, arsenic and fluoride results do not come with it.

Two operating lessons are portable. In Abu Dhabi’s pilot, dual-purpose wells showed indications of declining injection and abstraction capacity from clogging, while the infiltration basin was easier to operate and maintain, and the design acted on that.^12 3 During storage, 568 cubic metres a day is to be infiltrated continuously, and the modelling suggests that adjusting pumping on real-time monitoring could probably hold recovery efficiency at about 80 per cent rather than the modelled 60.³