1.

Introduction
Destructive earthquakes to date remain nature's most powerful display of force, which has
repeatedly caused severe devastation to both man-made structures and natural terrains.
(Vasquez Herrera & Dobry, 2019)
. Liquefaction is a geotechnical phenomenon that occurs when saturated
or partially saturated cohesionless soil loses its strength due to external stress, such as an
earthquake. As the ground shakes, the water pressure within the soil increases, causing the
particles to lose contact with one another and behave more like a liquid than a solid. This process
can result in the ground surface sinking, buildings tilting, and infrastructure collapsing,
underscoring the destructive potential of liquefaction in earthquake-prone areas
(Honegger & Wijewickreme, 2013)
. One of the most devastating examples of liquefaction flow failure is the
San Fernando Dam failure, which occurred due to the 6.6 magnitude earthquake on 9 February
1971 in Los Angeles, California. The dam experienced a slide on its upstream slope into the
reservoir due to excess pore pressure, which carried away a large portion of the core, leaving
only 4 to 5 feet of freeboard and causing severe cracks on the downstream slope (Seed, 1973).
Although a total catastrophic failure was avoided, the incident raised concerns about the existing
methodology due to its inability to predict such failures.

The failure of the San Fernando Dam during the 1971 San Fernando Earthquake has been
extensively studied, providing valuable insights into the behavior of embankment dams under
seismic loads. This event served as a critical turning point in the fields of geotechnical and
earthquake engineering, prompting researchers to investigate the mechanisms of failure and
propose improved design and maintenance practices.

Seed et al. (1975) were among the first to analyze the failure, identifying soil liquefaction as the
primary cause. This phenomenon caused the downstream slope of the San Fernando Dam to
collapse, leading to a partial breach. Seed and his colleagues emphasized the importance of soil
compaction and the role of pore water pressure in triggering instability. Finn et al. (1982)
extended this understanding by using numerical modeling to simulate the dynamic response of
the dam under earthquake-induced stresses. Their work highlighted the limitations of traditional
static stability analyses in predicting seismic performance. Chowdhury (Chowdhury et al., 2019;
Chowdhury, 2019) extensively studied the failure of the dam using the commercial software Fast
Lagrangian Analysis of Continuum (FLAC), which is based on the FDM. The Chowdhury
(2019) analysis was able to predict the initiation of the flow slide and the elevation of the
remaining freeboard at the heel scarp. However, the final post-earthquake geometry with runout
and blocky separation of the failure mass was not predicted due to the limitations of the FDM in
FLAC.

This study explores advancements in large-deformation analysis of earthen dam failures,

focusing on the failure of the San Fernando Dam. It highlights two approaches for analyzing
large deformations: (1) simplified liquefaction runout analysis using the Limit Equilibrium
Method (LEM) to evaluate pre- and post-earthquake stability, and (2) advanced numerical flow
slide runout analysis using the Material Point Method (MPM). The study adopts the latter,
showcasing MPM's ability to model complex failure characteristics.

The Material Point Method (MPM), originally developed by Sulsky et al. (1994), is a hybrid
particle-mesh numerical technique designed to address problems involving large deformations
and interactions between materials. It combines the advantages of Lagrangian and Eulerian
frameworks by representing material points with particles while computations are carried out on
a fixed background grid. Over the decades, MPM has evolved significantly and has been applied
to a wide range of engineering and scientific problems. Wang et al. (2016) introduced the
Generalized Interpolation Material Point (GIMP) method to reduce grid-crossing errors, a major
limitation of the original MPM. This enhancement significantly improved accuracy in problems
involving high gradients or sharp interfaces.

The study employed the Material Point Method (MPM) due to its unique advantages in modeling
complex large-deformation behaviors that are critical for understanding the failure mechanisms
of earthen dams like the Lower San Fernando Dam (LSFD). Unlike traditional methods, MPM is
particularly well-suited for simulating strain-induced degradation and post-failure runout
processes, which involve substantial material deformation and flow.

MPM allows for the accurate representation of complex failure characteristics, such as the
sliding and redistribution of soil masses, by overcoming challenges associated with mesh
distortion in conventional finite element or finite difference methods. This capability is
especially important for capturing the dynamic interactions between soil and fluid, including
hydrostatic pressures and reservoir-induced forces, during the runout phase.
Additionally, MPM facilitates the integration of advanced boundary conditions, such as
nonconforming Neumann stress boundaries and adhesive boundaries, which are critical for
realistic modeling of the reservoir pressure on the upstream face and interactions along sheared
planes. These features enable a more precise simulation of the transition from seismic loading to
post-shaking conditions, making MPM an ideal choice for this study.

2. Background
The San Fernando Dam located in the San Fernando Valley, served as a vital reservoir for the
Los Angeles Aqueduct system (David W. Sykora, 2019). Constructed by the Los Angeles Bureau
of Water Works and Supply in 1916, the dam was designed to provide water storage and flood
control for the growing population and agricultural demands. The reservoir was fed by a network
of small tributaries that were part of the dam's catchment area. By 1970, about 80% of the water
supply in Los Angeles came from the Upper and Lower San Fernando Reservoirs. Its failure
during the 1971 San Fernando Earthquake highlighted critical design flaws, geotechnical
challenges, and inadequate maintenance practices.

Geology and Foundation Condition

The embankment and abutments of the Lower San Francisco Dam were constructed on recent
alluvium, which consisted of stiff clay, sand, and gravel layers. This alluvium reached a
maximum thickness of 35 feet and was underlain by shales, siltstones, and sandstones. The left
abutment primarily consisted of Upper Miocene shales and siltstones, where the upper 30 to 50
feet (9 to 15 meters) was weathered and features numerous seams filled with gypsum. Excessive
seepage through the left abutment resulted from the dissolution of gypsum due to reservoir water.
The embankment of the Upper Dam was built on deposits of recent alluvium, which consist of
stiff clays and clayey gravels that are approximately 50 to 60 feet thick. However, the dam's
abutments are made of poorly cemented conglomeritic sandstone and coarse-grained sandstone
from the Saugus Formation.

Construction of Lower San Fernando Dam

The construction of the Lower San Fernando Dam started in 1912 and was built as a hydraulic
and rolling fill dam close to Granada Hills in the northeastern San Fernando Valley. The fill in
this method was collected using high-power water jets on a borrow area. The relative density of
the hydraulic fill ranged from 40% to 70%. Over the years, the dam was raised in stages, with the
final raise occurring in 1929-1930, bringing the crest height to an elevation of 1,144.6 feet. The
dam was 2,080 feet long and reached a maximum height of 142 feet with a crest width of 20 feet.
The downstream slopes were 2.5:1 (upper embankment) and 4.5:1 (berm), whereas the upstream
slope was 2.5:1. The last significant modification was made in 1940, with the addition of a rolled
earth berm at the downstream toe up to an elevation of 1096 feet to enhance the stability of the
dam. For many years, the dam operated with the Lower Van Norman Lake near full design
(Spillway) with an elevation of 1134.6 feet. However, in 1996, the highest operating reservoir
level was reduced by 9.6 feet to an Elevation of 1125 feet, and the spillway invert remained
unchanged (David W. Sykora, 2019; Seed, 1973)

Construction of Upper San Fernando Dam

In 1992, the semi-hydraulic fill method was used to construct the Upper San Fernando Dam,
which was situated north of the Lower San Fernando Dam and northwest of Los Angeles
(Seed, 1973)
. The original plan for the dam in 1922 was to build it to an elevation of 1238 feet (372 m).
However, it was ultimately raised to 1218 feet (371 m) by adding 50,000 cubic yards of
compacted dry fill on the upstream side. The finished dam crest measured 20 feet in width and
featured a 100 ft berm at an elevation of 1,200 ft, both the upstream and downstream slopes of
the dam were designed with a ratio of 2.5:1 (Seed, 1975).

Fig.1 The aerial view of the San Fernando Dam before the 1971 earthquake.
Seismic Stability Investigation Prior to Earthquake

A comprehensive assessment of the seismic stability of earth dams throughout California, done
in 1966, examined the earthquake resistance of the Lower San Fernando Dam using a typical
analytical method with a seismic coefficient of 0.15. This value was advised by a consultative
board designated by the Los Angeles Department of Water and Power, depending on the
established and anticipated seismic activity of the area.The strengths of the soils comprising the
embankment were determined by means of drained direct shear and triaxial compression tests on
undisturbed samples. The rate of loading in these tests may have been too fast for full drainage to
occur, but the data were interpreted conservatively to provide strength parameters for analysis
purposes. Stability computations were made using the conventional method of slices for the
combined effects of (1) An earthquake represented by a seismic coefficient of 0.15; and (2) a
partial drawdown/ of the reservoir level from El. 1,125-El. 1,110 (343 m-340 m). These
computations showed a minimum factor of safety

of 1.05. Based on the results of these studies, it was concluded that because the method of
analysis was based on conservative strength values and force applications in keeping with
conventional practice, the dam was safe against.

(JGED1 pdf)

Earthquake Damage to the Upper and Lower San Fernanao Dams in 1971

Damage to the Upper and Lower San Fernando Dams from the 1971 San Fernando earthquake
(Mw 6.6) helped scientists understand how these dams behaved seismically. Seismoscope data
recorded on the Lower Dam corroborated previous rock-site measurements showing peak ground
accelerations between 0.55 and 0.6g. A five-foot downstream displacement and a three-foot
settlement were among the many damages sustained by the Upper San Fernando Dam, which
also included longitudinal cracks and irreversible deformations. Piezometers detected significant
increases in pore water pressures; water levels rose by 8 feet at the downstream slope, 17 feet in
the dam's core, and 8.5 feet further downstream. These pressures may have caused the flow to
collapse if the safety factor after liquefaction had been less than 1. A disastrous collapse was
probably averted, nevertheless, by the dissipation of surplus pressures, particularly close to the
downstream slope.

Lower San Fernando Dam, on the other hand, saw a tremendous upstream fall that sent huge
chunks of its blanket into the reservoir, reducing the dam's freeboard to as low as five feet. Thirty
seconds after the end of the earthquake's shaking, the dam's static weight started the slide, which
continued for fifty seconds. 80,000 people living downstream were forced to evacuate their
homes and the reservoir level was lowered after inspections conducted after the earthquake
found a nearly vertical scarp and displaced debris. After the Upper Dam, piezometers
downstream measured three to five feet of higher water levels, but lower pore pressure ratios due
to the granite blanket's ability to dissipate pressure. Regardless, the upper dam's pore pressures
were probably similar to those at the dam's core. Critical lessons for seismic dam safety and
failure processes were provided by the disparities in damage amongst the dams, which
highlighted variances in material characteristics and pore pressure dissipation.(re-evaluation
pdf)

Aerial view of Lower dam failure

 Fig. Upstream face of the dame after drawing down reservoir

David W. Sykora. (2019). Case Study: Lower San Fernando Dam (California, 1971).

Honegger, D. G., & Wijewickreme, D. (2013). Seismic risk assessment for oil and gas pipelines.
Handbook of Seismic Risk Analysis and Management of Civil Infrastructure Systems, 682–715.
https://doi.org/10.1533/9780857098986.4.682

David W. Sykora. (2019). Lower San Fernando Dam (California, 1971) | Case Study | ASDSO
Lessons Learned. https://damfailures.org/case-study/lower-san-fernando-dam-california-1971/

Seed. (1973). Seed: Analysis of the slides in the San Fernando... - Google Scholar.
https://scholar.google.com/scholar_lookup?title=Analysis%20of%20the%20Slides%20in%20the
%20San%20Fernando%20Dams%20During%20the%20Earthquake%20of%20Feb.%209%2C
%201971%3A%20Technical%20Report%2073%E2%80%932&author=H.B.
%20Seed&publication_year=1973

Seed, H. B. , L. K. L. , I. I. M. and M. F. I. (1975). JGED1975 GT07 - Seed - The Slides in the San
Fernando Dams During the Earthquake of February 9,1971. ASCE, J of the Geotechnical
Engineering Division, GT7, Pp. 651-688.
Vasquez Herrera, A., & Dobry, R. (2019). Re-Evaluation of the Lower San Fernando Dam. Report 3.
The Behavior of Undrained Contractive Sand and Its Effect on Seismic Liquefaction Flow
Failures of Earth Structures. https://www.researchgate.net/publication/235202770