Lower San Fernando Dam — the Earthquake That Liquefied a Dam Above 80,000 People

Summary

At about six o'clock on the morning of 9 February 1971, the magnitude-6.6 San Fernando earthquake shook the Lower San Fernando Dam — a 142-foot (43-metre) earth embankment built by the hydraulic-fill method between 1912 and 1916 at the head of the San Fernando Valley in Los Angeles — and the upstream face of the dam slid bodily into the reservoir. The loose sand at the core of the embankment liquefied, lost almost all of its strength, and flowed. The crest dropped, the upstream shell spread some 250 feet beyond its original toe, and when the movement stopped only about five feet of soil — roughly a metre and a half — separated a full reservoir from the breach. No one died at the dam. Below it lay a residential district of some 80,000 people, and the cause was the seismic liquefaction of the dam's own fill.

The dam did not fail because the earthquake pushed it over. It failed because the shaking destroyed the strength of the soil holding it up. For about twelve seconds of strong motion, cyclic stress drove pore-water pressure in the saturated hydraulic-fill sand of the upstream shell until the effective stress between grains approached zero and the material behaved as a heavy liquid. Then — and this is the detail that made the case famous — the dam did not move during the shaking. The major slide occurred an estimated twenty to thirty seconds after the ground stopped moving, when the now-liquefied mass could no longer carry the dead weight of the embankment above it and the whole upstream slope ran out under static gravity alone.

Authorities did not know how close they had come until daylight. The Los Angeles Department of Water and Power began an emergency drawdown of the reservoir within hours and the police evacuated roughly 80,000 residents from the valley below while the lake was lowered over three days. The Lower dam held by a margin measured in feet. Had the reservoir stood a little higher, or the slide run a little farther, the case would read like Malpasset or Vaiont. Instead it became the most studied near-miss in geotechnical history.

The investigation, led by H. Bolton Seed and his colleagues at the University of California, Berkeley, and later re-examined by the US Army Corps of Engineers Waterways Experiment Station, established the modern understanding of how earthquakes destroy dams: not by inertia, but by liquefaction, and not necessarily during the shaking, but in the seconds after it. The finding ended the use of hydraulic fill for embankment dams in seismic regions and rebuilt the way every earth dam in earthquake country is analysed. The Lower San Fernando Dam is the canonical case of seismic soil liquefaction.

---

Timeline

1912–1916

Construction by hydraulic fill

The Los Angeles Bureau of Water Works and Supply builds the Lower San Fernando (Lower Van Norman) Dam to store Owens Valley aqueduct water. Soil is sluiced into place as a slurry, depositing the finest, loosest sand in a saturated, uncompacted central zone — the method that defines the dam's weakness.

1916–1930

In service and raised

The embankment is completed to a height of about 142 feet and a crest length of 2,080 feet, later modified and raised. It impounds the Van Norman reservoir directly above the densely populated north San Fernando Valley.

1930s–1960s

Decades without a strong shock

The dam performs for half a century under static loading and minor seismicity. Its hydraulic-fill core is never tested by strong ground motion, and the latent liquefaction hazard remains unexpressed and unmeasured.

9 Feb 1971, 06:00:55

San Fernando earthquake

A magnitude-6.6 shock ruptures on a thrust fault about 13 km from the dam. Peak ground acceleration at the site is estimated near 0.55–0.6 g, with roughly 12–15 seconds of strong shaking.

+~12 seconds

Pore pressure peaks

High pore-water pressures develop through an extensive zone of saturated hydraulic-fill sand in the upstream shell; effective stress collapses and the material liquefies.

+~20–30 seconds after shaking

The upstream slope slides

With shaking ended, the liquefied zone can no longer support the embankment's weight. The upstream slope and crest flow into the reservoir as a delayed static flow slide, spreading about 250 feet beyond the original toe.

9 Feb 1971, morning

Freeboard measured in feet

Inspection finds the crest dropped and the parapet carried into the water; only about five feet (~1.5 m) of soil separates the reservoir from overtopping the failed section. The reservoir stands at roughly El. 1109, holding about 10,000 acre-feet.

9–12 Feb 1971

Emergency drawdown and evacuation

The Los Angeles DWP draws the reservoir down rapidly while police evacuate approximately 80,000 downstream residents for three days until the lake is low enough to be safe.

1971–1972

Berkeley investigation

H. Bolton Seed, Kenneth Lee and colleagues study the slide. They conclude liquefaction of the hydraulic fill, not seismic inertia, caused a flow failure that moved after the shaking stopped — overturning the inertial 'pseudo-static' model of dam seismic safety.

1972 onward

Reservoir retired, Los Angeles Reservoir built

The Van Norman lakes are removed from full service; the lined Los Angeles Reservoir is constructed nearby to replace the storage, and the old hydraulic-fill dams are buttressed or decommissioned.

1988–1992

Corps re-evaluation

The US Army Corps of Engineers Waterways Experiment Station re-examines the slide using steady-state (residual) strength concepts, refining how the undrained strength of liquefied soil is measured and used in design.

The Build: A Dam Made by Sluicing Sand Into a Slurry

The Lower San Fernando Dam was built between 1912 and 1916 by hydraulic filling, a method that was efficient, cheap and, for a structure expected to survive an earthquake, almost ideally dangerous. In hydraulic fill, soil is mixed with water into a slurry and pumped or sluiced into the embankment, where it is allowed to settle out. The coarse particles drop near the outer shells; the finest sands and silts wash to the centre and settle slowly through standing water into a saturated, very loose, uncompacted mass. No rolling, no compaction, no control of density: the dam's strength is left to whatever the settling slurry happens to produce. The result at Lower San Fernando was a central and upstream zone of clean, loose, saturated fine sand — the single most liquefiable material a civil engineer can build with.

For static loading this was tolerable. A loose saturated sand under steady gravity drains slowly and holds, and for fifty-five years the dam stood without distress, raised over time to a height of about 142 feet with a crest some 2,080 feet long. It impounded the Van Norman reservoir at the northern edge of the San Fernando Valley, directly above neighbourhoods that by 1971 housed tens of thousands of people. The hazard below the dam was extreme and the hazard inside it was invisible.

The flaw was not a detail of the design; it was the design philosophy of the era. When the dam was built, the dominant model of earthquake resistance was the pseudo-static method: treat the shock as an extra horizontal force, a fraction of gravity pushing sideways, and check that the embankment will not slide as a rigid block under that force. That model says nothing about what cyclic loading does to the pore pressure inside saturated sand. It cannot see liquefaction, because liquefaction is not a force on the dam — it is the disappearance of the dam's strength. A structure designed and judged by the pseudo-static method could pass every check and still be a flow slide waiting for its earthquake. Lower San Fernando was exactly that structure.

The Failure Sequence: Twelve Seconds to Liquefy, Thirty More to Slide

At 06:00:55 on 9 February 1971 a magnitude-6.6 thrust earthquake ruptured roughly 13 kilometres from the dam. The site experienced strong shaking for about twelve to fifteen seconds with peak accelerations estimated near 0.55 to 0.6 g. As the ground reversed direction many times a second, the saturated hydraulic-fill sand in the upstream shell was sheared back and forth. With each cycle the loose grains tried to compact, but water filling the pore space could not drain in the time available, so the load shifted off the grain skeleton and onto the water. Pore-water pressure climbed cycle by cycle until, after about twelve seconds, it had risen across an extensive zone to nearly the total overburden stress. At that point the effective stress between grains fell toward zero. The sand had liquefied: a soil that a moment earlier could carry a dam was now, mechanically, a dense fluid.

Then came the detail that rewrote the discipline. The dam did not slide while the ground shook. The inertial forces of the earthquake, the thing the old method worried about, were not what moved it. The slide occurred an estimated twenty to thirty seconds after the strong motion stopped. With the shaking gone and no driving inertia at all, the liquefied zone simply could no longer hold up the weight above it. Under static gravity alone the upstream slope and the crest flowed downhill into the reservoir, the embankment spreading about 250 feet beyond its former toe. The crest dropped, the upstream parapet rode down into the lake, and the remaining intact section was reduced to roughly five feet of freeboard above a nearly full pool. The mechanism was a delayed flow slide: shaking does the damage by raising pore pressure, but the failure is a static collapse of a now-strengthless mass, and it can arrive after the danger seems to have passed.

It is the timing that made the dam survivable and the lesson permanent. Because the slide ran out toward the reservoir rather than downstream, and because the failed remnant happened to stop a metre and a half short of the water, the dam was not breached. Had the reservoir been higher, the failure would have overtopped and unzipped the embankment, and a wall of water would have descended on 80,000 sleeping people. The Lower San Fernando Dam did not pass its test. It failed and was saved by geometry and luck.

The Reckoning: Liquefaction Becomes a Discipline

The investigation was led by H. Bolton Seed and Kenneth Lee at Berkeley, with extensive sampling and cyclic laboratory testing of the hydraulic fill. Their conclusion was unambiguous and, at the time, radical: the dam had failed by liquefaction of its loose saturated sand, the failure was a flow slide driven by loss of strength rather than by seismic inertia, and it had moved after the shaking ceased. This overturned the pseudo-static model that had governed dam design for generations. The relevant question was no longer 'how hard does the earthquake push the dam?' but 'how much strength does the soil lose, and can what remains carry the structure?' That reframing — from inertial force to residual strength — is the intellectual core of modern geotechnical earthquake engineering, and it begins here.

The case was reopened in the late 1980s and early 1990s by the US Army Corps of Engineers Waterways Experiment Station, whose re-evaluation analysed the slide using steady-state, or residual, strength concepts. By measuring the undrained strength that liquefied sand retains once it is flowing, and correcting laboratory values for the void-ratio changes caused by sampling and by the 1971 shaking itself, the work refined how engineers quantify the post-liquefaction strength used in stability analysis. The Lower San Fernando slide thus generated not one but two foundational advances: the recognition of seismic flow liquefaction in the 1970s, and the steady-state strength framework for analysing it in the 1990s.

The practical verdict fell on the construction method itself. After 1971, hydraulic fill was effectively abandoned for embankment dams in seismic regions; no one would again build a water-retaining structure out of uncompacted, saturated, loose sand in earthquake country. The Van Norman reservoir system was retired from primary service and replaced by the lined Los Angeles Reservoir built nearby, and the old hydraulic-fill embankments were buttressed or decommissioned. The dam that nearly drowned 80,000 people became the reference case taught in every soil-dynamics course in the world.

---

Contributing Factors

01

Liquefiable hydraulic-fill construction

The dam was built by sluicing soil as a slurry, which deposited a loose, saturated, uncompacted zone of clean fine sand at its core — the most liquefiable material in civil engineering. No compaction control existed; the embankment's strength was left to chance settling. A modern rolled, compacted, well-drained fill would not have liquefied. The failure was latent in the construction method from the day the dam was finished.

02

Pore-pressure buildup under cyclic loading

Twelve seconds of strong shaking sheared the saturated sand repeatedly. Because the loose grains tried to densify but water could not drain in time, load transferred from the grain skeleton to the pore water, raising pore pressure to nearly the overburden stress. Effective stress collapsed to near zero and the soil liquefied. This is the mechanism the older design model could not represent, because it is a loss of strength, not an applied force.

03

A delayed, static flow slide

The dam did not move during the earthquake. The major slide came an estimated twenty to thirty seconds after shaking stopped, when the liquefied zone could no longer support the embankment under gravity alone. This decoupling of failure from shaking is the case's signature lesson: seismic inertia triggers the damage, but the collapse is a static event that can arrive after the ground is still.

04

Design by the pseudo-static method

The dam was judged by treating the earthquake as a steady horizontal force and checking rigid-block sliding. That method is blind to liquefaction because it never models pore pressure or strength loss in saturated sand. A structure could satisfy every pseudo-static check and remain a flow slide awaiting its earthquake. The analytical framework, not just the soil, was inadequate to the hazard.

05

Extreme downstream consequence with no margin

Roughly 80,000 people lived directly below a dam built of liquefiable sand, and the failure left only about five feet of freeboard between a full reservoir and a breach. There was no redundancy, no buttress, no margin against the governing failure mode — only the accident that the slide stopped short. The hazard rating of the population at risk was never matched by the seismic robustness of the embankment. ---

Aftermath

No one died at the Lower San Fernando Dam, but only because the slide halted about five feet below overtopping and the reservoir was drawn down before the next aftershock. Roughly 80,000 residents were evacuated for three days; the broader San Fernando earthquake killed about 64 people elsewhere and caused some $500 million in damage. The dam's technical legacy is enormous and specific. Hydraulic fill was effectively ended as a construction method for embankment dams in seismic regions, and every existing hydraulic-fill dam in earthquake country was flagged for re-evaluation. The pseudo-static method was displaced by liquefaction-based and deformation-based seismic analysis: engineers now estimate the pore-pressure rise, the post-liquefaction residual strength, and the resulting flow or deformation, rather than checking a rigid block against a fraction of gravity. The Berkeley investigation and the later Corps re-evaluation together established seismic soil liquefaction as a quantitative engineering discipline. The Van Norman system was replaced by the Los Angeles Reservoir, and the old dam became the byword for a single proposition: an earth dam can be destroyed by the strength its own soil loses, seconds after the shaking has already stopped.

---

Lessons

Never build a water-retaining embankment from loose, saturated, uncompacted sand in earthquake country — hydraulic fill is a flow slide waiting for its shock; compact the fill, control its density, and drain it.
Analyse the strength the soil loses, not only the force the earthquake applies — pseudo-static checks are blind to liquefaction; model pore-pressure rise and the residual strength that remains, and prove the structure stands on it.
Expect failure to arrive after the shaking stops — a delayed flow slide can run out under static gravity once the soil has liquefied, so a dam that survives the tremor intact is not yet proven safe.
Match seismic robustness to the population at risk — a dam above 80,000 people demands margin against the governing failure mode, not freeboard measured in feet and survival measured in luck.
Re-evaluate every old structure against the failure mode the original code could not see — the Lower dam passed the analysis of its era and still nearly breached; a method that cannot represent the hazard is not a defence against it. ---