Engineering Geology, 24 (1987) 339--353 339

Elsevier Science Publishers B.V., Amsterdam- Printed in The Netherlands

LATEST THINKING ON THE MALPASSET ACCIDENT

G. POST and D. BONAZZI

Coyne et Bellier, 5 rue d'H~liopolis, 7501 7 Paris (France)
(Accepted for publication December, 1986)

INTRODUCTION

After much thought, we have decided that there are two additional points
deserving discussion in connection with the causes of the Malpasset accident,
which supplement those already widely published in the literature.
The first point concerns the personalities involved in the design and
construction of the dam. This would be only a very minor addition to the
historical record if it did not involve what is still a very topical preoccupation
concerning engineering judgement and responsibility.
The other point is a second look at the Malpasset rock, more especially
the rock in the wedge which disappeared from the left bank, as seen and
described by the engineers involved in the design and construction work
before the accident, and which we can still try to grasp t o d a y from the
information in the records and through analytical techniques which did not
exist in the fifties.

ENGINEERING JUDGEMENT AND RESPONSIBILITY

A n d r ~ C o y n e and the design o f thin arch d a m s

Andr~ Coyne was a man of unlimited professional talent and personal

qualities, with great charisma. The group of young engineers in his firm in
the fifties were totally devoted to him. In the professional c o m m u n i t y at
large, he was highly respected both in France and abroad, as can be glimpsed
from the fact that he was to become President of ICOLD. Given this and his
personal charm, it was inevitable that he should be a leading voice in all the
circles where he was working at that time.
The decision that an arch dam was technically feasible at the Malpasset
site was thus based on his own personal conviction, which naturally carried
his colleagues along. Arch dams were something of a speciality with Andr~
Coyne. He had been the driving force which led to about 30 being built by
the time Malpasset was designed.
When reading his class lectures at the Ecole des Ponts et Chauss~es, one is
struck by the following sentences: " A type of structure which has never
collapsed is a rare and probably unique thing in engineering. But despite

0013-7952/87/$03.50 1987 Elsevier Science Publishers B.V.

340

appearances, despite its slender shape and elegant lines, and high stresses, it is
a fact that the arch dam is the safest of all structures. This is only added con-
firmation of what has been known for thousands of years about the stability
of arches". His lectures also contained a solemn warning that the abutments
are the most important part. Stated briefly, nothing serious can happen to an
arch dam " . . . provided the abutments stand up".
There was no way he could have imagined a failure scenario which would
have enabled him to go further in his analysis. Why should he have considered
the a b u t m e n t rock sliding along a set of planes which had not been com-
pletely catalogued at Malpasset? It was only afterwards that it became normal
practice to make a methodical inventory of the rock structure in depth. Nor
could he have conceived of the mechanical effects of water under pressure
seeping into the foundation through cracks caused by the dam itself just on
the upstream side of the abutments. Engineers' concern with these factors
t o d a y stems from the accident.
One imagines that the only mechanism he could have envisaged was mass
shear, and even poor rock can be expected to have a wide margin of safety
against this type of failure.

P e r m a n e n t technical c o m m i t t e e on d a m s

This brief outline of the situation at the time needs not be gone into at
any more length. It is, however, worth remembering one of the immediate
outcomes of the accident in France. In an area where the empirical approach
was and still is very important, it is not advisable for major decisions to be
taken b y one person, however competent he may be.
There was a movement in the sixties in France, one of whose leading
exponents was Professor Barbier, which eventually led to a Decree on June
13th, 1966 setting up the Permanent Technical Committee on Dams. This is
a b o d y of eight experienced specialists from various backgrounds who must
approve and review all dam designs more than 20 m high to be built in
France.
The terms of reference of the Committee leave it enough freedom to give
speedy approval to new ideas it considers safe, which would not be the case
if these were a set of written regulations. But one must not hide from the
fact that increasing the number of persons involved in any major decision
does have its negative aspects, in that innovative, bold designs have less chance
of finding acceptance.
This new controlling b o d y is therefore not a universal panacea, but it has
had the merit of exposing all dam problems involving public safety to criticism
and debate at an early stage, before the final decision is taken. The fact that
it has been welcomed by the profession is a sure sign that it meets a real need.
 341

MALPASSET ROCK AND ACCIDENT (Figs.l, 2, 3, 4)

The Londe method

The theory elaborated in the sixties, and again described at this meeting
b y its e p o n y m o u s author, is formally irrefutable. Its unique feature was its
three<timensional approach to this space problem. Uplift pressures, which by
their very nature are unpredictable and in a state o f constant change, can be
handled as variable parameters, so that their effect on stability can be
examined and discussed, and usually controlled b y drainage. The Londe
(1968) theory is ideal for examining conditions at Malpasset, from which it
sprang. We believe that this theory explains all or most of the accident. We
must stress this point, so that what follows will not be taken as a criticism but
rather as filling a gap, certainly not an attack on the argument which originally
grew o u t of a consensus first published in France in the review Travaux in
July 1967 in an article b y Jean Bellier.
The French engineering c o m m u n i t y quickly accepted the argument,
almost to a man. Marcel Mary, Chairman of the French Committee on Large
Dams, publicly supported it as soon as it appeared.
However, there were, and still are, some reservations. One of the purposes
o f this paper is to try to examine them and see if they can be in any way

Fig. 1. Malpasset Dam.

342

Fig.2. The right bank after failure.

enlightening. Most were expressed b y French engineers who had no part

whatsoever in the Malpasset project but who, like all dam engineers, were
acutely concerned with the lessons which could be drawn from the accident.
Therefore we shall describe what we feel may be missing from the most
widely accepted explanation of the Malpasset failure, which may perhaps not
give enough weight to the deformability of the rock in the process. This gap
in the argument m a y be due to the fact that this effect in the rock, i.e., its
deformability, raises the question of the subjectivity of assessments on rock
quality.
The accident of course influenced every engineer's judgement on rock
quality in assessing other sites.

Malpasset rock deformability

Information available
It is perhaps an oversimplification to distinguish only between the left-
and right-bank rock at Malpasset, as is usually done, and conclude that the
left bank was probably weaker than the right. In the left bank, we might also
distinguish between the rock left in place after the accident, which can still
be seen t o d a y (let us call it rock 1, Figs.3 and 4), and what was carried away
(which we shall call rock 2).
 343

Fig.3. The left bank after failure.

There are some discrepancies or even contradictions between the descrip-

tions of rock 2 in the excavation inspection reports and the conclusions from
the laboratory analysis of rock samples taken after the accident from rock 1,
summarized by Professor Habib. Jacking tests made after the accident, of
course concerned rock 1.
It is conceivable t h a t rock 2, forming the wedge which disappeared, might
have loosened and weathered more than the rest of the rock at the site (effect
of tectonics near downstream fault). There is no way now of checking this
possibility, but it does lead to some reservations regarding the rock 1
observations. The data available are as follows, in chronological order.

Site inspection reports (extracts). The rock mentioned on the left bank is of
course rock 2).
5/6/1952: Left-bank excavation
"(Under a thickness of 3.50 m ) , . . , there is then a blue, sound rock
(gneiss) containing some small brecciated zones. After going about 1 m into
this blue rock, a suitable bearing surface was found for the dam, despite the
presence of brecciated zones, which were small. If any large brecciated zones
were found, local treatment would be done, by removing the crushed rock
and backfilling with concrete.
344

Fig.4. T h e f o l i a t i o n plan o n t h e left b a n k .

In any event, there does not seem to be any advantage in excavating deeper
b y going more than 1 m deep into the blue r o c k . . . "
25/7/1952:
"I -- Excavations. The excavations for the plinth have reached an advanced
stage, the rock has been exposed over the whole area, at a much shallower
depth than estimated from the single borehole in the river. The rock head
was found at around El. 3 8 . 0 0 . . .
The exposed rock is often very weathered. It occurs as plates (gneiss)
which can be easily broken by hand. However, a hole about 1 m deep seems
to indicate an improvement of the rock in depth, which retains its plate
structure but becomes much harder.
The excavations will therefore have to be deepened by about 1 m to
1.50 m, which will set the b o t t o m line at around El. 36.50. The use of
explosives should be kept to a minimum, and n o t used for the last 50 cm.
 345

In addition, the rock seems to weather in contact with air and damp. The
surface must therefore not be left exposed to the air for any length of time
before being covered with concrete. The last 20 cm of rock can be left in
place, and removed at the last moment."
29/1/1953: I -- E x c a v a t i o n s . . .
"Left bank.
Between joints K and L, the rock did not improve in quality despite going
deeper But because the dam will be slotted into the rock, the foundation
can be considered acceptable.
Right bank.
For about 10 m on either side of joint F, the downstream half of the
excavated surface is a plane of smooth soft rock dipping upstream under a
layer of harder r o c k . . .
No attempt will be made to expose this smooth surface over the whole
width of the e x c a v a t i o n . . . The transverse step will simply be cut b a c k . . .
to give a somewhat smoother surface to the excavation."

1959 instrumental data from the dam. Fig.5 shows the dam deformations
measured from September 1955 to July 1959. An analysis with the
COQUEF3 computer program described later shows that there were substan-
tial displacements at the base of the dam.

1960 jacking tests. "Elastic deformation m o d u l u s . . . 850 MPa vertically

and of the order of 1,600 MPa horizontally (NS direction and EW direction)"
(Bellier, 1967).
"The deformation modulus, derived from the jacking t e s t s . . , or analysis
. . would be around 1000 MPa, perhaps even less. It is a well-known fact
that such conclusions must be approached with caution, but this does not
alter the fact that the modulus is low." {From Mary, 1968.)

COQUEF3 analysis. The COQUEF3 analysis is the first one of a set of

investigations recently undertaken by Coyne et Bellier to try to analyse the
part played by the rock deformability in causing the failure. Appendix I
describes the method and first results
The purpose of these initial simplified analyses was to check how far the
assumptions on the stiffness of the foundation based on the reports quoted
above agree with measured displacements.
The model was successfully calibrated, and confirmed the following points:
high concrete modulus of 25,000 MPa; moderate rock modulus at lower part
of foundation, 1000 MPa; lack of symmetry between the two banks, so that
the overall 1000 MPa modulus must be replaced by 1500 MPa for the right
bank and 500 MPa (with an artificially stiffened thrust block) for the left
bank.
There is of course the usual problem of making any accurate assessment of
foundation moduli because dam displacements are due mainly to concrete
deformations, even if the abutment deformations are unusually high in this
particular case.
346

ARCH~EL.98 ARCH AT EL.90 ARCH AT EL.78

N pO
/

'~ \4 L

/H

I' j.~O / JO

,B c ARCI-4._ATEL.66 ARCH ATEL.54

D
ARCH AT EL,45

- /

H H

--- F

(1) Scale
(b
(A)September 1955 level: 79.75 0 20 40 m m
(B) July 1956 level=83.85
(C)July 1958 level=87,30
(D)July 1959 level 94.10

Fig.5. Malpasset D a m - - plan views o f d e f o r m a t i o n before failure.

Conclusion. All these facts together mean that it is conceivable that the elastic
modulus of the left-bank rock was very low {500--1000 MPa), particularly
between joints K and N. If this is so, then the left-bank thrust block would
have been subject to much higher loads than it had been designed for, say
25,000 tonnes instead of 5000 tonnes, acting in a more unfavourable direc-
tion. One of the very first events in the process would then have been the
failure of this thrust block, because it had to withstand a major part of the
thrust from the dam towards the left-bank.

Other effects of deformability

Upstream crack. Such large left-bank deformations arising from the low
elastic modulus, which may or may not have caused the thrust block to start
sliding on its base (the 80 cm movement?) would be accompanied by the
heel crack widening very rapidly, from the value measured in July 1959,
which was approximately 15 mm at joint J at the bottom o f the left-bank.
 347

The arch thrust acted tangentially to the foliation on the left bank. One
can therefore suppose that the foundation sheared along the foliation plane,
as well as cracking and moving towards downstream. Shearing along the
foliation would probably widen the heel crack (cracking o f the foundation
just upstream o f the heel o f the dam has become a usual assumption since
the Malpasset accident) in a downstream direction along the dip (Fig.4) and
a further loading of the thrust block. The computer analyses now in progress
are trying to check these points.
Whatever the outcome, one must clearly associate large rock deformation
with wide heel cracking.
If the rock is naturally watertight and deformable like at Malpasset, there
is inevitably a large permeability contrast between the upstream and down-
stream sides of the dam, if only because of the heel cracking, regardless of
any gain in watertightness on the downstream side, through the Bernaix
effect.
Imagine two such zones of rock at Malpasset. The upstream zone is
decompressed with wide cracks at the foot o f the dam, while the downstream
zone is highly compressed and yet exposed to the hydrostatic pressure of the
reservoir. Even if there had been the usual drilled drains, it seems to be not
very likely that t h e y could have reduced the uplift forces under the dam.
In a word, high rock deformation probably makes drainage unfeasible. Yet
this is just a situation were drainage would be most necessary, because defor-
mable rocks are the weakest, and least capable of withstanding the extra
loads caused b y underseepage behind the dam. In hard rock, the dam forces
produce less of a permeability contrast so that the rock can drain naturally,
and in any event, there is a greater reserve of strength in the rock against
such effects.

Blow-out risk. As is shown by the computer analyses, low and varying elastic
moduli produce large deformations and tensile forces so that zones of highly
divided rock m a y be decompressed to the point where t h e y may be expelled
b y the hydrostatic forces within them. Under a suitable combination of
conditions, the rock m a y effectively be expelled locally. Even if only small
rock volumes are involved initially, the process will grow inevitably and
rapidly in such well-jointed and fissured conditions as at Malpasset, until,
eventually, the dam collapses. But there is no way of knowing when the first
blow-out actually occurred, or even, if it did occur, whether it was a necessary
or sufficient condition for failure.
On a more general level, it can be said that a highly deformable rock sub-
jected to arch thrust and seepage pressures undergoes a combined displace-
m e n t towards the banks {from the arch thrust) and in a downstream direction
(because of the heel cracking). It is the resultant of these combined effects
which we must consider.
348

HYPOTHESES ON MALPASSET FAILURE

It is repeated t h a t the foregoing discussion is not meant to be a substitute

for the process proposed by Londe and Bellier. It is quite true that the down-
stream fault was a potential cause of failure, and anyone looking at the site
t o d a y can immediately imagine the rock sliding along the downstream face
of the wedge.
But for this movement to start, there had to be a high water pressure in
this downstream fault, and the question arises as to whether this pressure
would have existed there if the rock had been less deformable, in other
words, if it had been better? Some d o u b t must subsist.
On this proposition, one is then justified in incriminating the poor quality
of the rock due to the finer divisions, as well as (and perhaps as much as) its
larger scale jointing and faulting. There seems no d o u b t that the final act in
the collapse process was mass-sliding of the left abutment, and this pos-
sibility has now become second nature to dam designers. But we feel that one
can learn more from this accident by stipulating a minimum acceptable
order~f-magnitude elastic modulus for rocks to be considered capable of
withstanding the thrust from an arch dam, if we suppose that the mass
movement was only the last act in a process which originally developed from
some other first cause.

TENTATIVE DAM SITE CLASSIFICATION ON THE BASIS OF ROCK

DEFORMABILITY

It is an accepted fact in t o d a y ' s rock mechanics that there is no overall

theoretical answer to the question of the stability of a rock a b u t m e n t sub-
jected to arch thrust. It is still a matter of experience and precedents.

Chart o f the Electricit~ de France

This is illustrated in Fig.6, which is a chart that the Geological and Geo-
technics Department of Electricit~ de France has been keeping for many
years. The moduli are deformation moduli, based on the total displacement
of the plate from the start of the jacking test.
Malpasset classifies as having the most deformable rock of all the sites,
some of which were considered more suitable for embankment dams or
straight or arched gravity dams. But in fact, if one discards the suspect
figures recorded at the test adit crown, the values are higher {2000 MPa
max., 1000 MPa mean, 400 MPa min.).

Heel cracking vs. dam height

Another chart is based on the postulate that upstream-to-downstream

displacements o f the foundation are o f the same order of magnitude on
average for the actual arch and for a vertical slice of unit width applying a
horizontal thrust o f 1/4h2 on the foundation, in which h is the dam height.
 349

MPA
20000- I ..:{

./
l 0 000

2000 ~::: e>.

/ :!
/
"<:;.;.-% ...../
!
1000 /
500 ::;] "<:.i
I
,-',--arch dam
I00 .... " "'1 i " "

Fig.6. Deformation moduli for dam foundations (jack test). After Electricit~ de France.

A two-dimensional FEM analysis gives an order of magnitude for the

width F o f the heel crack versus the elastic modulus of the rock E and the
d a m height.
These orders of magnitude are shown on Fig.7 with points representing a
n u m b e r o f other French dams. The moduli are taken from jacking tests or
back~analysis on recorded deformations. They agree quite well with the
values from the 1Ah 2 formula. The chart is far from complete, however, and
further data could well be added in the future.
Malpasset lies right on the left o f this chart, with a heel crack width of
2 cm. It is tempting to plot a boundary on this chart for sites suitable for
arch dams, although it would be somewhat arbitrary. Rather than a single
line, therefore, it would be better to have a strip dividing sites unsuitable for
arch dams from those where an arch dam is obviously feasible, separated by
an area where the final choice depends on other factors. This has n o t been
done on Fig.7. It is submitted to the engineering c o m m u n i t y for scrutiny
and discussion.

CONCLUSION

The Malpasset accident suddenly made dam engineers aware that water
pressures in the foundation had to be controlled for arch dams as well as
gravity dams, and the effects of certain discontinuities in the left a b u t m e n t
made t h e m realize that the internal structure of the rock had to be investi-
gated before applying enormous loads to it.
Furthermore, while the stability o f the left a b u t m e n t was no d o u b t
jeopardized b y the major discontinuities, which were well documented, a
factor was probably also the gradual deformations caused b y the finely
divided nature of some parts of the foundation. Various measurable param-
eters (like jacking tests) n o w yield information on this point.
350

height ofdom
I
(rn}
200 /

50 50 A ' LE GAGE

150
i;o/ B : LAOUZAS
C : LAPARAN
ID , M A L P A S S E T I
E = Ste CROIX
F = VOUGLANS
G : MONTEYNARD
I00 .. H , TIGNES
j_ J

50

4" / f /

, I_..
6 10 15 20 ErlO~lPa

Fig.7. Tentative relationship between upstream crack vs. dam height and rock modulus.

The feasibility study for an arch dam must therefore henceforth include
detailed investigations guided b y a geologist who is familiar with arch dam
design as well as the local and regional geological context, accompanied b y
in-situ and laboratory tests, e.g., core drilling, permeability tests, jacking
tests in adits, Petite Sismique survey, effect of water on compression test
results, etc.
All the surfaces of discontinuity that are representative for the structure
of the abutments, as used in the Londe method, must be explored. Attention
must also be given to rock deformability, and it seems reasonable to stipulate
some minimum elastic modulus for a r o c k suitable for an arch dam. We would
suggest a minimum value o f 3000--6000 MPa for arch dams of low or
moderate height, more for very high arch dams. The final choice should be
governed b y other factors such as depth of excavation compared to dam
height, and whether or not the grout curtain will join up with a concrete
apron covering the zone affected b y heel cracking, or some other t y p e of
rock improvement. The sensitivity o f the modulus of the rock after saturation
must also be examined.
Such investigations will classify sites into whether they are acceptable,
marginal or unacceptable for an arch dam. Experienced engineers can imme-
diately recognize the two extreme cases without awaiting test results, b u t
testing and classification charts are useful for intermediate cases. The final
decision will still be based on experience and judgement, and it is important
for engineering judgement to be exercised early, freely and, if possible, after
frank discussions between specialists in the various areas involved.
 351

APPENDIX I -- 1985-ANALYSIS

Method

T h e t w o steps in t h e analysis were as follows.

Calibration. T h e m o d e l was " c a l i b r a t e d " o n t h e basis o f a c t u a l d i s p l a c e m e n t s m e a s u r e d
o n t h e d o w n s t r e a m d a m face at t h e following dates: S e p t e m b e r 1 9 5 5 , reservoir level
7 9 . 7 5 ; J u l y 1 9 5 6 , reservoir level 8 3 . 8 5 ; J u l y 1 9 5 8 , reservoir level 8 7 . 3 0 ; J u l y 1959,
reservoir level 9 4 . 1 0 .
T h e s e r e c o r d s are a l m o s t ideal, b e c a u s e all f o u r sets o f m e a s u r e m e n t s were m a d e in
s u m m e r so t h a t seasonal t e m p e r a t u r e effects m u s t be small.

I CQYNE ET BELLIER I t COQEF3

I [ Fiq. 8

MEASURED .... 0 1 2cm

DISPLACEMENTS
COMPUTED -

EL. 83.85

I I E L . 87. :50
I I

/ I
/--4
f i
/ I
EL.94.10

J I
t---4
I
t----.t

NOTA: ZERO DISPLACEMENT AT ELEVATION 79.75

0 IOta

MALPASSET DAM JOINT H

EB=25000. ERR: 1500 ERL = 5 0 0 (MPA)

Fig.8. J o i n t F d i s p l a c e m e n t s .
352

COYNE ET 8ELLIER I COQEF3

I I Fig. 9

MEASURED .... 0 1 2m
DISPLACEMENTS
COMPUTED

L 85.85

87.30

/ L 94.10
I ~7 s

I-1

\
\

NOTA: ZERO DISPLACEMENT AT ELEVATION 79.75

0 20 m

MALPASSET DAM ARCH AT ELEVATION 79.50

EB=25000, ERR=1500 ERL=500 (MPA)

Fig.9. D i s p l a c e m e n t s o f t h e arch at e l e v a t i o n 79.50.

T h e p u r p o s e o f c a l i b r a t i o n was t o find s u i t a b l e elastic m o d u l i for t h e c o n c r e t e and

f o u n d a t i o n , a n d used a simplified m o d e l in w h i c h t h e d a m was r e p r e s e n t e d b y t h i c k shell
e l e m e n t s a n d t h e f o u n d a t i o n , b y Vogt springs.
T h r e e - d i m e n s i o n a l non-linear analysis. T h e s e c o n d step ( n o w in progress) uses a finer
m o d e l o f t h e d a m a n d f o u n d a t i o n , c o m p l e t e w i t h its geological singularities. T h e u p s t r e a m
a n d d o w n s t r e a m faces o f t h e r o c k w e d g e are m o d e l l e d b y n o n - l i n e a r joints, a n d t h e
c o n c r e t e , a n d some o f t h e f o u n d a t i o n rock, is m o d e l l e d w i t h a n o - t e n s i o n material.
T h e j o i n t s are m o d e l l e d as elastoplastic unilateral ( o p e n i n g ) j o i n t s with, if possible,
C o u l o m b limit friction.
Uplift pressures are gradually i n t r o d u c e d into t h e j o i n t s as t h e y o p e n .
 353

Calibration

Arch dam geometry. The dam (a Pascal spiral) is represented with its actual geometry,
on the actual foundation surface. The left-bank thrust block is represented by thickening
the thick shell elements.
Foundation model. Semi-infinite elastic space using Vogt elements.
Steps. Three models were prepared, with different assumptions for the Young modulus
for the concrete end rock.
Model 1: Econcrete = 10 000 MPa Erock = 1000 MPa
These values were taken from a three-parameter fitting calculation in 1960. This model is
not included in the calibration process.
Model 2: Econc~ete = 20 000 MPa Emck = 1000 MPa
The results from this model indicate that the true concrete modulus is higher than
20 000 MPa. Model 2 is symmetrical and does not reproduce the slight asymmetry of
rock behaviour towards the left bank. It does, however, demonstrate that the Ec/E r ratio
is higher than 20.
Model 3: Econcrete = 25 000 MPa
ERBmck = 1500 MPa
ELB rock = 500 MPa (joints K to P)
The thrust block foundation was artificially stiffened (to 10 000 MPa to account for the
restraining effect of the foundation excavation and of the wing wall).
Results o f calibration. The match between measured and computed displacements is
shown in Figs.8 and 9. Origin of displacements is reservoir level 79.75. Measured and
computed displacements are shown at El. 83.85, 87.30 and 94.10.
Model 3: Displacements. The match is excellent on the right bank, but displacements
seem to be restrained on the left bank.
The movement is more towards the abutment than towards downstream at the top.
The downstream movement of the thrust block is not reproduced.
The match is excellent at the bottom.
At the base of the dam (joint H, El. 45), displacements are slightly underestimated by
the computation (35% approx.).

REFERENCES

Bellier, J., 1967. Le barrage de Malpasset. Travaux.

Londe, P., 1976. Proceedings of the Conference on Evaluation of Dam Safety. Pacific
Grove, Calif., Nov. 28--Dec. 3, 1976.
Mary, M., 1968. Barrages-Vofites, historique, accidents et incidents. Dunod, Paris.