Landslide

By David N.Petley, Ph.D.

View of landslide scarp looking toward South with landslide mass in foreground.

he remarkable Vajont reservoir disaster of Northern Italy

is a salient example of an attempt to apply engineering
measures to a poorly understood but highly hazardous
system. The disaster occurred when approximately 270 million
m3 of material o n one of the banks of a newly-constructed
reservoir failed catastrophically during the filling. The movement was truly catastrophic, with movement rates estimated
to be as high as 30 m/sec. The landslide displaced a wave that
overtopped the dam by up to 250 m and swept onto the valley
below, causing the deaths of more than 2,500 people. Remarkably, the dam remained unbroken by the flood.
This article presents background information regarding
the dam and its geologic setting, the sequence of events
leading to the failure, causes of the failure, and some of the
important lessons learned from this tragedy.

BACKGROUND
Vajont is located in the southeastern part of the Dolomite Region of the Italian Alps, about 100 km north of

Venice (Figure 1). The dam was built as a part of the postwar development of Italy to provide hydroelectric power for
the rapidly-expanding northern cities of Milan and Turin.
While siting a dam at this location was first proposed in
the 1920Js, construction did not begin until 1956, When
completed in 1960, the doubly-curved arch dam was, at
265.5 m above h e valley floor, the world's highest thin arch
dam. The chord of the dam was 160 m, and the volume of
impounded water was 115 million m3.
The dam was built across the Vajont Valley, a deep, narrow gorge. In this area, the mountains are characterized by
massive, near-vertical cliffs formed in the Jurassic Dogger
formation and underlying Triassic formations. The valleys
are generally formed by synclinal geological structures that
expose weaker Cretaceous and Tertiary units, which contain
more clays and are more thinly bedded (Figure 2). Hence,
the valley walls are formed from middle Jurassic limestone,
overlain with successive layers of upper Jurassic limestone
with clay and Cretaceous limestones.

--

Limit of 1960 landslide

0

0.5

1.0

from Jaegar 1980

ua

I
I

Figure 1 . Map showing location of Vajont Dam and Reservoir.

ORDER OF EVENTS
Before Dam Completion
It appears that during construction of the dam in 1958 and 1959,
engineers were concerned about the
stability of the banks of the reservoir. As a result, a possible prehistoric slide was identified o n the right
bank. While there was considerable
discussion of the stability of the
valley walls in view of the inclined
synclinal form of the strata and the
possibility of old slides in this area,
deep-seated landslides were deemed
extremely unlikely because:
areas of weakness were not identified in the test borings (although it
appears that only three boreholes
had been drilled in the reservoir
walls);
a potential shear plane was
assumed to have a 'chair-like' form
that would resist movement;
seismic analyses suggested that the
banks consisted of very firm, insitu rock with a high modulus of
elasticity.
Smaller slides in the looser surface

layers were considered to be likely, but

the volumes and velocities of movement were expected to be low.

First Reservoir Filling and

Drawdown
Filling was initiated in February
1960 (Figure 3) before final completion of the dam in September of that
year. A month after filling started,
when the reservoir level had reached
130 m above the river level, a small
slope movement occurred o n the
south bank. As reservoir filling continued, engineers monitored the
movement of the banks. In October
1960, when the depth of the reservoir
had reached 170 m, they observed a
rapid increase in the rate of displacement to approximately 3.5 cm/day.
At the same time, a 2-km length joint
opened, defining an area of about
1700-m long and 1000-m wide, suggesting that a very large landslide had
been mobilized. On November 4, with
the reservoir depth at 180 m, a failure
of a part of the landslide mass, consisting of about 700,000 m3, occurred
over a period of about 10 minutes. In
response, the reservoir level was gen-

tly drawn down, whereupon rates of

movement of the remaining body of
the landslide decreased from a maximum of about 8 mm/day at a level
of 185 m to less than 1 mm/day at a
level of 135 m. The total displacement
of this main landslide body was about
1 m.
Site engineers realized that a large
mass of the left bank was inherently
unstable and prone to creep. They
decided to control the movement of
the sliding mass by varying the reservoir's water level, while controlling
the joint water thrust within the rock
mass by means of drainage tunnels.
The designers realized that a section
of the reservoir could be blocked by
the landslide mass, but the volume of
water in the unblocked (upstream) section would be sufficient to allow the
generation of electricity. Therefore, a
bypass tunnel was constructed on the
opposite (right) bank such that if the
reservoir was divided into two sections
by a landslide of the left bank, the level
of the lake could still be controlled.
The engineers assumed that by
carefully altering the level of the reservoir, they could initiate and control

-_- - . . --_ Yr_ .=_T.-0-I:<.~-

. . --

:-5

~- -

- c - -

=
7
.

---

- -c--*-2

-- . -.- ? - r T - - - s - : ?=-?LC '-f

Y 5

z-

15.

x=21?-

- 7 -- -- --

--~.
r3 =.-.-----L
-I=-:
:
I
.:
:7---I

- -1;
. *-.
.
....
--. *- - - 7 .T7
7
;
>
-?
?
:<
:.
:--.
. ..
. - --=
:
5
--LT.
a
L
?
!
C
?.me. ~---- - - - - 5 - - - . _c -_- - - - - --. .--x u.&ety
A

Second Resemoir Filling

and D ra~s-d
own
' r ~ z 1 :he beginning of October

- :... cnul early February 1962, the

-:
-

-.carer level was raised to 185 m, followed by a phase of slow impoundment. By November 1962, the level
had reached 235 m. During the early
part of this phase, slape velocities did
not substantially increase; but by the
end of the phase, they had increased
to 1.2 cm/day.
In November 1962, the dam's engineers slowly lowered the reservoir
level a second time, until the water
decreased to 185 m after four months.
Initially, displacements remained high,
but in December they began to reduce
and, by early April, the rate was effectively zero. The experiences gained
from the second phase of filling and
the subsequent drawdown confirmed
to the engineers that control of the
landslide was possible by altering the
level of the reservoir.

Third Reservoir Filling and

Drawdown
Between April and May 1963, the
reservoir level was rapidly increased to
231 m. Slight increases in velocity were
noted, but rates did not exceed 3 mm/
day. In June, the level was increased
to 237 m and the rate of displacement increased to about 4 mm/day.
In mid-July, the level reached 240 rn,
and some of the control points indicated small increases in displacement
to about 5 mm/day. This level was
maintained through mid-August, but
during this time, velocities steadily
increased to 8 mmlday. In the latter.
part of August, the level was increased
once more, and by early September,
the water depth was 245 rn. In some
parts of the slide, velocities increased

Self-Anchoring, Multi-Purpose Platform

20-Ton Push Capability Fully Operable Using Wireless Remote
Hydraulic Chuck *Track Mounted
Equipped with Percussion Hammer for Soil & Groundwater Sampling

to as much as 35 mm/day.
In late September, the water level
was slowly dropped t o bring the
rates of creep back under control.
By October 9, a depth of 235 rn was
reached, but velocities of movement
continued to slowly increase, and
eventually rates of u p to 200 mm/
day were recorded.
March / A p r i l 2006

Catastrophic Failure
The catastrophic failure of the
landslide took place at 22:38 GMT
on October 9, 1963. The entire mass
slid approximately 500 m northwards at up to 30 rn/sec. The mass
completely blocked the gorge to a
depth of up to 400 m, and traveled

29

from Jaegar 1980

!
L

Figure 2. Schematic cross-section through Vajont Valley.

u p to 140 m up the opposite bank

The landslide's movement ceased
after about 4 5 seconds. At the time
of failure, tbe reservoi~.contained 115
million m 3 of water. The landslide
mass pushed a wave of wawr up the
opposite bank that destroyed t h e
village of Casso-260 m above lake
level-and over-topped the dam by
u p to 245 m The water, est~matrdto
have had a volume of about 30 million m3, then cascaded more than
500 m down the valley to destrov the
villages of Longarone, Pirago, Villanova, Rivalta and Fae, causing about
2,500 fatalities. Surprisingly, the dam
was not destroyed.

CAUSES OF FAILURE

from Jaegar 1980

--

tlgure 3 Chronology of events leading to the cata5trophic failure

Since 1963, Illany studies have

been undertaken t o iderltify t h e
causes of the failure. Initially, there
was considerable speculation about
the location of t h r sliding surface,
but it is now established that it was
located in thin (5-15 cm), over-ronsolidated clay layers in the limestone.
To date, experts debate whether the
landslide was caused by reactivation
of an old landslide or by a first-time
failure. It is now, however, firmly
established from detailed analyses of
the movement record that the final
Failure occurred in a brittle manner,
with the rapid acceleration resulti n g from t h e sharp reduction i n

shear resistance as the clay weakened

from peak to residual strength, and
k o m the steep shear surface. When
r h e reservoir level was high, the pore
pressures in the clay were elevated,
reducing the resistance t o shear
stress. The rapid drawdown had the
effect of inducing hydraulic forces in
the limestones, which increased the
overall shear stress as water drained
from the slope. However, the lower
permeability of the clay meant that
pore pressures remained high in
these materials. Thus, the slope experienced an increase in shear stress at
the same time as shear resistance was
minimal. The catastrophic failure was
the result.

LESSONS LEARNED
While debate continues over the
details of what happened within the
slope at Vajont in the period leading u p to failure, there are a number of important lessons that can be
learned from the tragedy. Perhaps the
most important lesson is the dangers
posed by attempts to control complex
systems when the mechanisms that
control them are poorly understood
and the consequences of failure are
severe. Viewed from a 21st century
perspective, the attempts to control
the landslide through changes in
lake level seem somewhat unwise,
given that the detailed processes of
failure were, and in many ways still
are, poorly understood. Secondly, the
failure emphasizes the importance of
understanding the actual conditions
in the slope through detailed ground
investigations. Thirdly, it starkly illustrates the threat posed by large-scale
failures in reservoir banks-hopefully, those involved in the very large
dam projects of today are fully aware
of these threats.
Finally, it is now well-established that
simple extrapolations of the increase
in movement rate would have allowed
prediction of the time of final failure of
the slopes some 30 days or more before
the failure actually occurred. While it is
harsh to criticize those working on the
site for their failure to do this (given
that these techniques were not really

Presto's Free Technical Resources Library CD has the tods

you need for solving soil stabilization problems and porous
pavement needs with the Geoweb" and Geoblock@systems.

Environmentally-Sound Solutions - Load support,

slope and channel protection, earth retention and

understood at that time), hopefully,

engineers faang a similar challenge in
the future will use the movement record
to the fullest extent possible to extrapolate likely future behavior. 0
Dr. David N. Petley is Wilson kofessor of Hazard and Risk in the Department of Geography at the University of

Durham, located in Durham, UK. His

research interests focus on understanding
landslides and landslide mechanics, especially in high mountain areas within less
developed countries. To this end, he established and is director of the International
Landslide Centre (wuno.landslidecen h-e.
org) Dr. Petley can be contacted at d.n.
petley@durharn.ac.uk.