Who "stole" the Danube water?

Along the Romanian reach, stretches of riverbed that usually lie beneath the Danube have become broad banks of pale sand. Boats have settled onto dry ground in Serbia, while rocks and shoals have emerged around bridge piers in Budapest. The low water has disrupted freight and river cruises. It has also constrained cooling water for nuclear power stations in Romania and Hungary. These are tangible consequences, recorded in several countries along the same river [1]. Anyone looking at the narrowed channel in Romania is entitled to ask where the water went.

The measurement at Romania's upstream boundary makes the crisis stark. Between 7 am on 3 August and 7 am on 4 August 2026, the National Institute of Hydrology and Water Management recorded a discharge of 1,450 cubic metres per second at Baziaș, where the Danube enters Romania. The multiannual average for August is 3,900 cubic metres per second. In other words, the river was arriving with about 37 per cent of its usual August flow. The institute expected it to fall to 1,400 cubic metres per second the following day [2]. Concern is proportionate to the evidence.

Discharge and water level are related, though they are not interchangeable. Discharge measures the volume passing a cross-section each second. The height visible against a bank or gauge also depends on the channel's shape, sediment, local engineering and the operation of nearby structures. A shoal can become exposed after a relatively small fall in level where the river is broad and shallow. Conversely, a deeper reach may remain navigable even while less water passes through. The Baziaș discharge is therefore more useful for testing a basin-scale theft claim than a single photograph of a beach. It describes how much water crossed Romania's upstream threshold, rather than how deep the river looked at one location.

Online, that concern quickly acquired culprits. One widely circulated Romanian story claimed that powerful interests upstream were keeping the Danube in their own lakes. Different versions shifted responsibility among western countries, Ukraine and vaguely defined foreign actors. Reporting on the viral posts captured both the allegation and the way commenters supplied whichever country they already distrusted [3]. The details could change because the story's real appeal lay elsewhere: a frightening physical change had been given an intentional cause and a human enemy.

A conspiracy claim about a river can be tested against the river. Baziaș is especially useful because it is the checkpoint at Romania's entrance. Every shortage produced by low rainfall in Germany, reduced Alpine snowmelt, dry tributaries, abstraction in the middle basin or temporary upstream regulation is already reflected in the water measured there. The gauge cannot by itself separate all those influences. It does establish that the low flow was present before the Danube crossed into Romania. Any proposed explanation must fit that fact.

Following the river upstream also corrects a misleading mental picture. The Danube is more than a pipe beginning at a spring in Germany. Its basin covers over 800,000 square kilometres and extends across 19 countries when tributary catchments are included [4]. Water reaches the main stem through a connected network of rivers and groundwater across much of a continent. What crosses Baziaș is therefore the accumulated result of precipitation, snowmelt, evaporation, soil and groundwater storage, tributary inflow and human withdrawals throughout the upstream basin.

Evidence of dryness was already visible near the top of that system. On 16 July, the International Commission for the Protection of the Danube River reported persistent drought and little prolonged rainfall in the German part of the basin. Heat had raised water demand while replenishment of rivers and groundwater had weakened. Munich, in the Isar catchment that drains into the Danube, imposed restrictions on several uses of mains water and on withdrawals from rivers and lakes [5]. The supposed country hoarding the river was itself limiting local consumption because water was scarce.

River discharge integrates conditions over time. Rain that falls on one Romanian town does not instantly restore a continental river, particularly if much of the basin remains dry. Water may first wet depleted soil. Some evaporates or is taken up by vegetation. Some enters groundwater, then reaches a tributary slowly. Snow stored in the Alps matters in a different season. Tributaries respond at different speeds, and a brief storm can raise a local level without repairing months of deficient inflow. Travel time adds another delay: a change in an upper reach takes days to propagate along the main stem, while inflow from a distant tributary follows its own route. Hydrologists consequently compare measurements across stations and over successive days. They do not infer the state of an 800,000-square-kilometre basin from rainfall at one window. This is why a wet afternoon beside the lower Danube cannot refute drought upstream.

The wider pattern matters more than any one photograph. By late July, Budapest had recorded an exceptionally low level, cargo traffic had largely halted there, and boats were stranded near Novi Sad. Hungary's meteorological service reported a substantial 90-day precipitation deficit across nearly the whole country [1]. Romania's gauge then showed the same depleted river arriving at Baziaș. These observations are separated by borders and gathered by different institutions. Together they trace a low-water event along the river's direction of travel.

Hydropower is the most plausible-sounding part of the hoarding story. Dams are visible, governments regulate them, and reservoirs plainly can store water. Yet the type and operation of a plant matter. Uniper describes its 13 power stations on the German Danube as run-of-river plants. Its control centre manages flow and level across the barrages so that locks and navigation can function [6]. VERBUND likewise identifies four power stations on the Upper Austrian Danube as run-of-river facilities [7]. These are different from large seasonal reservoirs designed to accumulate an enormous spring inflow for release months later.

A run-of-river plant generates electricity from water moving through a modest height difference. Water is guided through turbines and returns to the river downstream. A barrage may create a headpond and operators can adjust gates, so the plant can influence water level locally and shift some flow over hours. During maintenance, flood management or navigation operations, those adjustments matter. Water can also bypass the turbines through spillways when inflow exceeds generating capacity. In each route, it continues downstream. Limited storage cannot make the continuing inflow disappear. Sustained generation depends on water continuing downstream, and operators cannot repeatedly refill a headpond without new water arriving from above. When the river weakens, these plants lose generating potential too. They depend on the same depleted flow and should not be imagined as continental-scale taps.

That explanation should not be turned into a defence of every dam. Engineering has profoundly changed the Danube. The ICPDR identifies hydropower, navigation works, flood protection and water abstraction as important pressures on the river's physical condition [8]. Dams can interrupt fish migration and slow sediment transport. Channels can redirect much of the flow away from an original reach. Dykes sever floodplains from the river, while peaking operations can produce abrupt fluctuations. The Iron Gates system, jointly operated by Romania and Serbia, retains sediment and has altered flow velocity and habitats across a long impounded reach [9]. These are measured environmental effects with specific mechanisms.

Abstraction also deserves scrutiny. Farms, cities and industry remove water, and the combined demand becomes more consequential during drought. Good water management therefore requires public data on withdrawals, reservoir levels and environmental flows. Still, the viral allegation needs evidence of an extraordinary upstream intervention large enough to explain the observed deficit. A sustained withholding of 1,000 cubic metres each second would accumulate 86.4 million cubic metres in one day. Reservoir levels, operating records and downstream gauges would register an action of that scale. A suspicion is not converted into hydrology merely by attaching it to a dam.

The Bâstroe Canal illustrates why precision matters. Work on the navigation route in the Ukrainian part of the Danube Delta has raised legitimate transboundary concerns. The documented issues include dredging, suspended sediment, habitat disturbance and changes to local water-level dynamics. International bodies have called for environmental assessment and cooperation between Romania and Ukraine [10]. Those concerns deserve serious monitoring. They do not explain a low discharge at Baziaș, which lies far upstream of the Delta. A project near the river's mouth cannot cause the river to have already arrived depleted at Romania's western entrance.

The same distinction applies throughout the basin. A structure may cause major ecological damage in a bypassed channel while returning most of the diverted water to the main river farther downstream. A reservoir may change the timing of flow without consuming the stored water. Irrigation can consume water through crop use and evaporation, while a turbine does not. Sediment can be trapped even as water passes through. Collapsing all of these processes into the phrase “they kept our water” erases the very evidence needed to decide which impacts are real.

International governance supplies routes for investigating real disputes. The Danube River Protection Convention commits the basin's states to sustainable and equitable management of surface water and groundwater, with cooperation on transboundary effects [11]. A treaty does not guarantee perfect compliance, and it should never substitute for measurements. It does mean that extraordinary claims can be tested through shared monitoring, operating data and formal challenge. If an upstream state were withholding a consequential volume, the responsible response would be to identify the facility, quantify the missing flow and compare gauges above and below it.

Climate change belongs in the account with similar care. The wider consequences of climate change for public health are explored elsewhere in Research Explained. The current heat and prolonged dryness are consistent with a warmer Europe in which evaporation intensifies and drought risk grows. Basin authorities already plan for more frequent water scarcity [5]. Warming can increase the atmosphere's demand for moisture and alter how much winter precipitation is stored as snow, yet the effects on a particular river episode depend on when and where precipitation failed. Land use, soil condition and water demand can intensify the outcome. A projection about future risk is different from an attribution study of this particular event. Without a dedicated analysis comparing the observed episode with a counterfactual climate, it would be too strong to calculate how much of the 2026 low flow was caused by human-driven climate change. The immediate observations establish severe dryness across the basin. Longer-term climate science explains why low-water planning has become increasingly urgent.

The Danube has not been untouched by human power, and Romania has sound reasons to demand transparent management upstream and at home. Dams alter ecosystems. Withdrawals compete for a finite flow. Delta engineering can carry transboundary risks. Those truths make the conspiracy more persuasive because they lend it real pieces of river history. The decisive evidence in August 2026 remains simpler: the Danube was already exceptionally low when it entered Romania, after persistent drought across the basin. Misinformation did not invent the crisis. It took a visible, frightening shortage and attached the most emotionally satisfying component: the idea that someone who is “against us” must be to blame.