How a Spillway Eats Itself

Water has a strange failure mode. Push it fast enough over a solid surface and the pressure in the fast-moving layer can fall until the water flashes to vapour, forming tiny cavities in the flow. Sweep those cavities a little downstream, into slower water at higher pressure, and they collapse, each one imploding with a shock that hammers whatever surface it collapses against. This is cavitation, and on a high-velocity spillway it is the most destructive process an engineer designs against, because it is not the flood that does the damage. It is the water attacking the concrete from inside its own flow.

Glen Canyon Dam in Arizona keeps two 41-foot concrete-lined tunnel spillways, one in each sandstone abutment. In June 1983 a very large Colorado River snowmelt forced them into real use for the first time. The left tunnel ran for seventy-two hours at 20,000 cubic feet per second, and when engineers looked, cavitation had already stripped about fifty cubic yards of concrete from the invert at the vertical bend. The trigger was almost nothing: a calcite deposit a quarter of an inch high, about six millimetres, on the tunnel floor. In flow moving on the order of thirty metres a second, that bump was enough to drop the local pressure into cavitation and begin tearing the liner apart.

But the reservoir was still rising toward the top of the gates, so there was no option to stop. The damaged tunnel had to keep running, and on 28 June, as the release was pushed higher, the flow gave the crew a signal no one could miss.

the water downstream from the spillway bucket turned an ominous amber color as the jet carried chunks of concrete and sandstone into the river.

Burgi, Moyes and Gamble, "Operation of Glen Canyon Dam Spillways, Summer 1983."

When the flow was finally shut off and the tunnels dewatered, the scale of it was clear. At the base of the vertical bend the left tunnel held a hole thirty-five feet deep, a hundred and thirty-four feet long and fifty feet wide, cut clean through the concrete liner and down into the sandstone, with three-quarters of the tunnel-liner circumference gone at the worst section. Roughly three hundred cubic yards of concrete, rock and steel had been carried down into the flip bucket, and a single sandstone boulder measuring eight by fifteen by fifteen feet had come to rest in the tunnel invert about five hundred feet short of it.

Reinforcement steel which looked like "spaghetti" was observed extending from the damaged tunnel liner in the horizontal section of the right spillway.

Burgi, Moyes and Gamble, "Operation of Glen Canyon Dam Spillways, Summer 1983."

What fixed it was not thicker or stronger concrete, because no concrete survives sustained cavitation at those velocities. It was air. During the repairs, Reclamation cut aeration slots into each tunnel, ramps that lift the jet and draw a layer of air in beneath it. A few percent of air along the invert makes the flow compressible enough to cushion the imploding cavities, and the shock never reaches the concrete. When the tunnels ran again after the retrofit, the cavitation did not return.

The lesson: On a high-velocity spillway, tolerance is measured in millimetres. A joint, an offset, a form line, a calcite bump, any of them can drop the local pressure into cavitation and start a runaway that a single flood can finish. Grind the invert flush. And accept that past a certain velocity, surface finish alone will not save you: the cavitation index falls too low, and aeration stops being an upgrade and becomes the only reliable defence. Cavitation is not abrasion. It is not sediment wearing the surface down; it is the water boiling and imploding against it, and the cure is not hardness but air.

Read more: Spillway Concrete: Abrasion and Cavitation Resistance →