The Trend That Didn't Reverse

A big concrete dam is never still, and that is exactly as it should be. It leans a little downstream as the reservoir fills and swings back as the reservoir empties; it expands in summer and contracts in winter; its crest traces much the same loop year after year. All of that movement is reversible and predictable, and a dam is wired to measure it: piezometers for the uplift pressure under its base, weirs for the seepage through it, thermometers for its internal temperature, and plumb lines and surveys for its deformation. The job of surveillance is not to find a dam that has moved. Every dam has moved. It is to find the small part of the movement that does not come back.

The Zeuzier dam, a concrete arch in the Swiss Alps first filled in the late 1950s, taught the profession how much that distinction is worth. For two decades its behaviour was unremarkable. Then, late in 1978, its pendulums began reporting a movement that would not fit. It was not the reversible swing of reservoir and temperature; it was a slow, one-way drift that kept growing, and because it decoupled from the two things that normally explain an arch dam's motion, it read as an alarm rather than noise.

In 1978, with reservoir full it started to deflect upstream, cracks up to 1.5 cm wide appeared on the downstream face parallel to both abutments and the radial contraction joints opened in the central part on the upstream face.

G.S. Sarkaria, "Lessons from Serious Incidents at Seven Arch Dams."

The cause was not in the dam, or even near it.

Investigations showed that the left abutment was moving towards the dam and squeezing it to destruction.

G.S. Sarkaria, "Lessons from Serious Incidents at Seven Arch Dams."

What was squeezing the abutment sat more than a kilometre away and some four hundred metres below the riverbed: an exploratory gallery, bored for a planned road tunnel, had drained the saturated, fissured rock, and as the water left, a basin of ground two to three kilometres across slowly settled. The Swiss federal regulator recorded thirteen centimetres of settlement at the dam and seven to eight centimetres of closure between the valley flanks. The reservoir was drawn down, the mechanism was diagnosed, and the dam was repaired and eventually returned to service.

The lesson: An instrument's value is almost never in a single reading. It is in the trend, and in how that trend compares to the behaviour you predicted. Zeuzier was caught not because the pendulums showed movement, but because the movement was irreversible and would not correlate with reservoir level or temperature, the two signals that explain a healthy arch dam. So read every instrument against a model of expected behaviour, strip out the reversible seasonal and reservoir response, and watch the residual. And keep the net wide, because the thing that destabilises a dam may be, as it was here, a tunnel no one on the dam-safety team was watching, a kilometre away and off every drawing. In India the Dam Safety Act now makes instrumentation and surveillance a statutory duty, not a courtesy. The dam will tell you it is in trouble. It tells you in millimetres, and only if someone is reading.

Read more: Instrumenting Mass Concrete: What to Embed and Why →