Under pressure: why we need to be thinking about dams

Dams rarely make the news, and that is exactly how it should be. Quietly, they supply the water that keeps homes, farms and factories running, and often absorb the shock of floods and droughts before those shocks reach people. Lately they have taken on yet another job. Semiconductor plants and the data centres behind artificial intelligence are thirsty facilities, and the industrys water use doubled between 2012 and 2022. The dams that hold that water now matter to national economies in a way they did not twenty years ago.

The impact of failure Which is why when a dam fails, the impact spreads far beyond the structure itself. In February 2017, for example, the spillway of Oroville Dam in California fractured under the strain of heavy water releases, and the emergency spillway that was meant to back it up began to erode. Around 188,000 people were evacuated. Nobody died, but the incident showed that even a well-resourced operator can be caught out, and that safety depends on more than the dam body itself: spillways, discharge systems and real-time operational judgement all matter. Two years later at Brumadinho in Brazil, a tailings dam, built to hold mining waste rather than water, collapsed without an effective warning and killed 270 people. The two failures had little in common technically, but they teach the same thing. By the time something visibly goes wrong, the options have already narrowed. The key to true dam safety is in detecting the warning signs early, not recovering well afterwards.

The worlds dams, meanwhile, are all ageing. According to a UN University report, most of the roughly 58,700 large dams registered with the International Commission on Large Dams (ICOLD) were built between 1930 and 1970, with a design life of 50 to 100 years, and by 2050 most people on Earth will live downstream of large dams operating at or beyond that design life. Climate change compounds the problem from the other direction. A UN World Water Development Report notes that per capita renewable water resources fell by about 20% between 2000 and 2018, while extreme floods and droughts arrive more often and hit harder. Structures designed for the last centurys climate are being asked to perform in this one.

Yet in most places, the way we check on these ageing structures has barely changed. Inspections are periodic and labour-driven. Workers walk the dam and climb its faces, and divers are sent down when something needs a closer look. Measurements are prone to gaps, drawings are stored on paper, and leak detection still depends heavily on what a person can see on site. This approach captures a moment, not a trend. It can respond to problems after they appear but it is poor at predicting them, and prediction is what ageing dams demand.

Bridging the gaps This is the gap that K-water, Koreas public water agency, set out to close. Between 2020 and 2025 it built a digital-twin-based safety management system, run by its Dam Safety Management Centre, across all 37 national dams it operates - twenty of them multipurpose, fourteen for water supply and three for flood control - and the system entered full operation this year. The idea is simple. Give every dam a set of senses that never blink and give its managers one place where everything that those senses report can be seen, compared and tested.

It works like this: GPS sensors and inclinometers track signs of deformation in the dam continuously. Thermal cameras scan concrete dams for surface leakage, while electrical resistivity surveys look for seepage inside embankment dams that are built of earth and rock, revealing the kind of internal warning sign that was never identified in Brumadinho. Aerial drones replace rope-access work at height, and underwater remotely operated vehicles take over the jobs that once required divers. An AI model then reads the drone imagery, flags suspected damage and helps generate the inspection report.

The digital twin is where it all comes together. Each dam site was captured with drones and radar at five-centimetre resolution and rebuilt as a three-dimensional model, layered with the building information model of the structure, live sensor readings, panoramic views of internal spaces, inspection and repair histories, and simulation tools for earthquakes and discharge scenarios. The result is less a display than a working environment for decisions. When an anomaly appears, an operator can see where it sits in the structure, what the sensors nearby have been doing, what was repaired there before, and what a proposed response would mean, all before committing crews or budget. Decisions get rehearsed as simulations first.

From reaction to prevention K-waters operational data makes the case plainly. When monitoring moved from manual and semi-automatic readings to real-time measurement, the rate of missing data fell from 32% to 3%. Drone-based aerial inspection cut inspection time by 90% and cost by 95% compared with conventional methods; underwater inspection times fell by 66% and costs by 85%. The point of those numbers is not saved budget. Cheaper and faster inspection means inspection can happen more often and reach further, and it takes people out of dangerous places. What changes is the posture of the whole organization, from reacting to what has already happened to anticipating what might.

So the technology works, and it is running at national scale. Whether it can travel is a different question, and it is mostly a question about standards. None of the building blocks involved are exotic. The three-dimensional geospatial data follows GML, the markup language standardized as ISO 19136. The building information models follow the IFC format of ISO 16739. The vocabulary of digital twins themselves has been standardized jointly by IEC and ISO as ISO/IEC 30173, drone operations are covered by the ISO 21384 series, and IEC

LINK:	https://www.iec.chUnder pressure: why we need to be thinking about dam...
	See more stories from iec

Most recent headlines