Failure of the St. Francis Dam
William Mulholland's Catesrophic Career-Ending Project
Shit happens. Things fail. Sometimes they are small things, but other times they are big things that affect lots of folks, usually in a bad way.
I love studying engineering failures, as we learn so much from our mistakes, and you can learn a lot about engineering decisions, tradeoffs, and even disasters. You almost never hear when things go right, or almost wrong, but you (ideally) get lots of details when things go terribly awry.
There are many good examples to write about, and I’ll do that this fall, but today, let’s look at one of my favorites, the St. Francisco Dam, and its catastrophic 1928 failure that killed at least 431 Californians.
Mulholland
Irishman William Mulholland, of Mulholland-drive fame, was first and foremost a civil engineer, spending much of his career focused on bringing drinking water to Los Angeles. You could argue LA would be nothing like it is today without his 50-year effort in the early 20th century to quench the thirst of the rapidly-growing city.
For references, Los Angles began the 20th century with 100,000 people, but by 1920 it was already nearly six times larger and would continue to grow rapidly for decades. Given it’s in a desert with poor and unevent rainfall, you can imagine the water supply challenges, then and now.
As head of the Los Angeles Water Department, Mulholland drove and supervised the building of the Los Angeles Aqueduct and a variety of other related projects from 1880 until the dam’s failure in 1928 ended his career and forever tarnished his reputation.
Along the way, it didn’t help that he had both extreme confidence and an ego to match, hubris that frankly contributed to the loss of 431 lives, the second-greatest loss of life in California's history (after the 1906 San Francisco earthquake).
In the end, the St. Francis Dam failed just 12 hours after he inspected it and pronounced it safe.
The Damned Dam
You can argue about whether or not his drive and success to water LA and fuel its growth was a good idea, but the St. Francis dam was doomed from the start, based on both where it was built, and how it was designed and built. Several interrelated mistakes and failures led, in the end, to its failure, and Mulholland’s, too.
Designed and built in the late 1920s when more and more water was needed, the site seemed ideal, as the choke point of the San Francisquito Creek and its eponymous canyon, 35 miles northwest of Los Angeles.
The large reservoir there would protect LA from interruptions in the LA Aqueduct due to earthquakes or other issues such as extended drought and, believe it or not, ‘terrorists’ who were blowing up the aqueduct on a regular basis.
Sadly, even though Mulholland thought the local geography and soils or rocks to be suitable for a large dam, they very much were not, a fact only ‘discovered’ after the dam failed.
Dam design at the time was evolving quickly, and this was only the second concrete dam Los Angeles had designed, the other being the nearly identical Mulholland Dam still standing (with reduced water loads) in Hollywood.
As a result of this design evolution and limited understanding of key stability factors of large concrete dams, several key elements were not included in the design, any one of which might have saved the dam.
Some of these were just lack of understanding of relevant physics and theories, including understanding of uplift forces and wells, arch dam forces, how water gets into and lightens concrete, along with how concrete should be made and placed in large structures.
Other mistakes were more questionable, such as taking the dam’s design from contemporary textbooks without any additional calculation nor analyses of the relevant site or forces.
Most puzzling was why the dam was built with a narrower base and smaller backside angles than were shown on the original permit filings - there were mysteriously never actually built in the field. Both were critical to the dams later instability, especially when its height was raised during construction.
The reality was the dam was fundamentally unstable as soon as it was built.
Geological Problems
In addition to the design problems, the Earth wasn’t very cooperative, either.
For starters, an earthquake fault ran under the dam, something you might think best avoided at any time, even though it was, and is, inactive.
The fault is there because it forms the seam between two very different rock formations, one of which is mica schist, which was later shown to be very unstable (Mulholland himself had concerns about it a decade earlier, but these were seemingly lost). The other side was red conglomerate, another less-than-desirable foundation.
Both of these rock types were later found to be very unsuitable for dams, as the conglomerate swells when filled with water, just as the schist can become unstable when saturated - both of which likely pushed upwards on the dam once it was filled, leading to the initial failure.
Much more importantly, the east side of the dam is an ancient landslide, and thus much less stable than solid rock formations. This would have been hard to detect in the 1920s, and was only discovered after the dam’s failure (and later found on 500+ other dams in the USA). In the end, it was a landslide in this area that likely triggered the dam’s final destruction around midnight on March 12, 1928.
I’ll Raise You
The dam was planned at 175 feet tall, with a deep base and ‘toe’ out in front to ensure it could hold back the water without tipping over. The design also included an increased back angle (batting) on the backside to reduce tipping forces.
However, Mulholland added 10 more feet as construction began, but without changing the rest of the design, such as the obvious move of increasing the dam’s base size. This reduced the overall safety margins.
Unfortunately, during construction, the toe was shorted 20 feet and the backangle never increased at the bottom. Both of these mysterious changes seriously reduced the dam’s ability to withstand the water pressure without tipping over.
Using modern dam analyses, engineers have shown that, as built and when filled to the top, the dam was actually unstable and easy to tip over. Not at good starting point.
My Crack or Yours?
A variety of cracks and leaks appeared soon after construction, but they were judged by Mulholland to be of no consequence, and it was considered a ‘dry’ dam. It remained so through 1927, but there was not a lot of water behind it due to the ongoing California Water Wars and related terrorist damage to the LA Aqueduct.
These wars resulted in a lot of water being drawn off the St. Francis reservoir, to be refilled later. Thus the stress on the dam was mostly low during this period and everyone thought they had a good dam. Until it refilled.
During this refill over the winter of 1927-28, stresses increased again and more cracks were found, this time larger and crossing both sides of the dam from their hillside abutments. Some of these were quite sizable, several inches wide in places.
Unfortunately, Mulholland made a major mistake in how these cracks were filled, sealing them from the front instead of the backside, which likely helped ensure the dam’s later failure. Sealing from the front ensured water continued to enter from the rear, saturating the concrete, and lowing the ability of the dam to withstand the forces on it, especially uplift.
The reservoir continued filling and more and worse leaks appeared, all the way into March, 1928, when the dam was filled to the top for the first time.
Finally being full caused more leaks, and on March 12, Mulholland again inspected these new and larger leaks, shown above, including one that was ebbing and flowing, perhaps indicating movement within the dam. Mulholland thought this should be fixed in the future and returned to Los Angeles.
It would be the last time he saw the dam intact.
The Failure
No one was there just before midnight on March 12, 1928, and there were no cameras, but we do have a few key pieces of information, from passers-by and from a water depth recorder mounted on the dam.
There was not a lot of early consensus on the failure, not even on which side of the dam the failure started at, nor why. It turns out there were enough problems to go around, between the west hills rock softness, the east hills instability, and the poor foundation design all around. But consensus later emerged, somewhat confirmed by modern modeling and surveying techniques.
The first known element was when a family drove by at 8:30 pm and had to briefly stop, as the road as a 12” vertical separation in it, indicating the land was already moving downhill a bit. This was the beginning of a land shift and likely increased pressure on the dam. The cause of this initial shift is unknown, but may have been water seeping into the hillside, lubricating it, for it was mica schist, and prone to sliding.
Around 11:30 the dam seems to have tilted upward and forward several inches. The core cause is unknown but likely it was already very unstable and the small landslide movements were enough to increase upward pressure sufficiently to fracture the weak concrete - the cracks from this fracture were visible on the backside of the remaining pieces after the collapse.
This fracture, crack, and tilting allowed water to enter the dam further destabilizing it. This setup a very dangerous situation, and probably caused further cracking and even front-wall fractures and high-pressure water leaks (as mist was reported downstream about this time).
This also probably changed the interaction of the dam and the eastern hillside, setting the stage for a much larger landslide that would destroy the dam. The time was about 11:40 pm.
A carpenter rode his motorcycle by the dam at 11:50 pm and heard a sizable landslide over his engine noise. This was the last minutes of the dam’s life, as over a million tons of rock started moving, dumping itself in front of, and behind, the dam, and also removing much of the dam’s east-side support.
Seven minutes later, at 11:57:30 pm, there was a power failure as power lines on the east hills fell in the continuing landslide (which don’t move very fast). Two minutes later the dam itself failed, as the landslide seems to have destroyed the left side of the dam, pushing the huge concrete blocks into the center of the river of now gushing water.
By 11:59 pm the 12 billion gallons of water behind the damn started surging down San Francisquito Canyon towards Ventura and the Pacific Ocean, 52 miles away, at 12 million gallons PER SECOND. The force was so great that it pushed the pieces of the dam downstream; the largest piece, weighing 10,000 tons, was pushed nearly a mile.
This is what the end result looked like, with a single piece still standing - the east side with the landslide is on the right side of this photograph:
This photograph is looking the other way, with the remaining damn piece in the middle of the lower left. The east-side landslide is clearly visible to the left of the dam area. The surge wave was 120+ feet deep going down this canyon:
The 120-foot high wall of water wound its way down to the Ocean, destroying power plants, houses, and whole towns as it went including 84 men who died when their construction camp was overrun.
The Aftermath
The dam was never rebuilt, and instead, the Bouquet Reservoir was constructed a couple of canyons east as a similar-sized holding reservoir, using a safer earth-filled dam. That is still in use today by the people of Los Angeles.
All federal dams were immediately inspected, as were all 32 of LA’s remaining dams, and 827 other California dams. About a third needed modifications. Later, over 500 dams in the USA would be found to have been built on previous landslides.
Additional rules were created to supervise dam design and construction, along with legislation to certify and license civil engineers, as this was far from the only engineering mess that caused loss of life in California.
The much larger Hoover Dam was built soon after, and the St. Francis engineering failures and lessons were heavily used, including moving the dam to a different site, reducing its loads, and increasing spillway and other related capacities. Perhaps we can thank the St. Francisco dam for a successful and stable Hoover Dam.
You can visit the St. Francis dam site, and perhaps I’ll make it up there someday, but there’s not much left and it’s a bit of hike as the old road was damaged by the landslide and abandoned. You can find it easily on Google Maps, and here is a video visit (in two parts).
But the pieces still remain there to remind us of engineering hubris and the lack of both safety factors and a solid design process with reviews, all necessary when and where our work affects the lives of others.
In the future, I’ll likely write about two more interesting failures:
The Ocrean Ranger Drilling Platform Loss in 1982 off Newfoundland
The Piper Alpha Disaster in 1988 in the North Sea