The entry says that the engineers read the slide for twenty years, and that in 1985 two of them put the filling record beside the rain. This piece is about that report: The Vaiont Slide, a Geotechnical Analysis Based on New Geologic Observations of the Failure Surface, by Alfred J. Hendron Jr. of the University of Illinois and Franklin D. Patton of Vancouver, written under contract for the US Army Corps of Engineers and published by its Waterways Experiment Station at Vicksburg in June 1985, in two volumes. It is a United States government work, and the one full account of the Vajont in English.
Why they went back
The report opens with the reason the Corps paid for it. “The large volume and high velocity of the Vaiont Slide combined with the great destruction and loss of life that occurred make it a key precedent landslide, particularly for slides caused by reservoir filling.” Dam projects in Mexico, Taiwan and Canada had been delayed or altered because of it; between 1964 and 1967 new regulations on reservoir slopes had come in in France, Germany, Italy, Japan and the United States. “Engineers and geologists are now generally obliged to examine the slopes of proposed reservoirs for the owners”, and when they find an old slide they have to explain why it is safer than the Vajont’s. “If the engineers cannot give a reasonably complete and consistent explanation of the Vaiont Slide, in terms of the currently available methods of stability analyses, then it is difficult to see how they can feel confident about their evaluation of other reservoir slopes. The disturbing aspect of previous reviews of the Vaiont Slide is that there are gross inconsistencies when the field data, slide behavior, and the results of analyses are compared.”
The inconsistencies were these. More had been published on the Vajont “than for any other slide in the world”, and yet no analysis had taken into account “(a) the obvious three-dimensional shape of the slide surface, (b) the actual laboratory shear strengths from representative samples of the material on the slide surface, and (c) reasonable piezometric levels related to both rainfall records and reservoir levels”. And “many authors have claimed or accepted the claims of others that there were no significant clays or clayey units present along the failure surface”. Müller, SADE’s consultant and the author of the principal papers in English, “made a point of dismissing the influence of clay interbeds”, writing that “clay and loam, however, were not present in the stratification joints of the Mount Toc”. So the authors went to the mountain. Edoardo Semenza, whose 1960 survey with Giudici had first called the slope an old landslide, went with them and gave them his maps, his sections and his own summary report of 1965, which they translated and printed as Appendix G; Donald Anderson did the calculations; samples went to laboratories in Edmonton, Urbana and Vicksburg.
Clay, and an old slide
They found the clay. Along the surface the slide had left bare, at outcrop after outcrop numbered on a photomosaic of the slope, they recorded “multiple layers of weak clays”, a few centimetres thick, “continuous over large areas of the surface of sliding”, in the beds at the base of the Lower Cretaceous and in the Malm below it; clay on the surface, clay below it, clay forming the matrix of the lower part of the slid mass, “thick clay fragments and layers” in the debris, and the same clay in the same beds outside the slide. “The authors find the field evidence for the presence of clay along the surface of sliding compelling because of the number of locations where clays were noted on the failure surface and because of the details of the geology at these locations.” The clay was mostly calcium montmorillonite. In the laboratory its residual angle of shearing resistance, the friction left after it has already been sheared, was 5 to 16 degrees, most of it 6 to 10; allowing for the places where rock touched rock they took 12 degrees for the base of the slide. The Frattini commission of 1964 had seen it too, and the report quotes it: “in the material accumulated by the slide we can see clay beds, a few centimeters thick … In our opinion, these strata, of a really clayey nature, cannot be considered the product of sliding; they may rather be of sedimentary origin.”
They found the old slide. “The 1963 Vaiont Slide was a reactivation of an old slide. The age of the old slide is unknown, but it probably occurred in postglacial times but before the period of recorded history of the Vaiont Valley. The evidence for an old slide is strong and diverse”: the drainage and the closed hollows on the slope, the breccias that had filled and refilled a crack at the head, the basal rupture surface, the remnants of a slid mass on the north side of the valley, which Giudici and Semenza had described in 1960. And they found that the “seat” of the slide, drawn as nearly horizontal in every cross-section, in fact dipped east, upstream, at 9 to 22 degrees, which made a two-dimensional analysis wrong and made about forty per cent of the resistance of the whole slope depend on the near-vertical faces of its eastern boundary.
They found the water. Above the slide, near the top of Monte Toc, was karst: sinkholes and solution cavities into which rain and snowmelt could run, and beds dipping towards the slide, so that water went in at the top and came out, under pressure, along the clay. “This evidence strongly suggests that the conditions were present to enable the transmission of high water pressures developed due to infiltration from precipitation or snowmelt on the mountain above.” Piezometer P2, when it was first read in 1961, stood 90 metres above the reservoir; its tip was probably not well sealed, so the real difference was greater; it rose and fell with the rain as well as with the lake. The clays and the limestones together made “a near classic case of an inclined multiple-layer artesian aquifer system at and below the surface of sliding”.
Rain beside the reservoir
Part VI of the report is the one that bears on the experiment. It takes the four years of the reservoir, 1960 to 1963, and sets the level of the lake, the rate of movement of the slope, the readings of the piezometers and the rain at Erto in ten-day totals one above the other on a single sheet, Figure 29, which the entry reproduces.
The first filling, from February 1960, produced the small slide of March at 595 metres, the acceleration of October at about 650, the perimeter crack, and the slide of 4 November; the movement reached 3 to 4 centimetres a day at the end of October, and October 1960 was the wettest month of the four years, with about 500 millimetres at Erto. The lake was lowered to 600 and the slope stopped. The second filling, from October 1961, went back through 650 in January 1962 with a velocity below a tenth of a centimetre a day, “in sharp contrast to the 3.5 cm/day velocity observed when the reservoir was just below el 650 m during the first filling”, and on to 700 in November 1962 with 0.2 to 0.3; then, with the lake held at 700, the rate rose abruptly to 1.2 “although the reservoir remained nearly constant”, after “a period of record precipitation for the four-year period”. The lake was lowered to 650 and the slope stopped, having moved 130 centimetres more. The third filling, from April 1963, reached 705 in July with 0.4 to 0.5 centimetres a day, 710 in early September with an immediate rise to 1.0, then 2 to 4 in the first days of October, and 20 on the 9th.
Then the authors state the hypothesis they are testing. “If other factors governing stability remain constant, then it is reasonable to assume that similar rates of movement should be observed for similar reservoir levels.” They quote Müller’s account of 1964 of how the engineers had read the second filling: the movements had a higher velocity “only if a new portion was wetted for the first time”, and that “led authorities and technicians to the conviction that a gradual stabilization of the moving mass would be brought about by raising the water level in individual steps”. And they answer it. “The erroneous assumption which led to the conclusions quoted above was that all other factors were remaining constant and the reservoir level was the main variable controlling the stability of the slide. In fact, rainfall was significant and was not remaining constant.” Every major movement, October 1960, November 1962, October 1963, followed heavy rain; wherever the same level gave different speeds, “the higher movement rate correlates with a higher precipitation rate”. The second arrival at 650, in January 1962, moved nothing “because of the low precipitation at that time and in the preceding months”; June 1963 at 700 moved at half the rate of November 1962 at 700, because June was dry. “Thus, it would appear from an evaluation of the records on Figure 29 that rainfall is as important as the reservoir level in determining the rate of movement of the slide.” No special mechanism, “creep” or “thixotropy”, was needed, “because the behavior only appears to be anomalous when the movements are correlated with reservoir levels without considering precipitation”.
Two tables make the point in numbers. In the fifteen days before the movement of 31 October 1960, with the lake at 645 metres, 170 millimetres of rain fell at Erto; in the fifteen days before 15 December 1961, when the lake came back to 645 and nothing moved, 50. In the fifteen days before 30 November 1962, with the lake near 700 and the slope moving 1.2 centimetres a day, 94.3 millimetres; before 30 June 1963, at the same level and almost still, 29.8. “For these examples, it is clear that the 15-day precipitation immediately preceding a significant movement at a given reservoir elevation was greater than three times the 15-day precipitation preceding the dates when the reservoir was at the same elevation but when no significant movements were observed.” The rain needed to move the slope also fell as the lake rose: 170 millimetres in fifteen days at 645 metres, 81 to 94 at 700.
Then they drew the envelope. On four charts, Figures 30 to 33, the level of the lake is plotted against the rain of the preceding 7, 15, 30 and 45 days, each reading dated, with the rate of movement in brackets. The points where the slope accelerated past half a centimetre a day lie along a curve; the quiet points lie below it. The curve is the “failure envelope”, “those combinations of water level and precipitation required for the slope to become unstable”. Extended to its ends it says two things. The reservoir level at which the slope would have failed with no rain at all was “approximately 710 to 720 m”, against a design full-supply level of 722.5; “had the reservoir been filled to its design level, the slide might have moved during a period without any significant preceeding rainfall”. And the rain that would have moved it with no reservoir was about 700 millimetres in thirty days, which, since nearly 500 fell in one month of the four years, “has been exceeded during the postglacial life of the slope. Therefore, significant movements must have occurred without a reservoir.” That, they add, “would appear to provide quantitative verification of the stories attributed to the local inhabitants about the occurrence of slope movements from time-to-time”. The name of the mountain, Semenza notes in his appendix, means “crazy” in the dialect.
The stability calculations, done in three dimensions with the measured clay strengths and water pressures tied to rain and lake, gave factors of safety that fell to about one at each of the four periods of movement and rose between them, with only small variations in the parameters possible. Filling the reservoir alone reduced the factor of safety by about 12 per cent; the assumed increase in pore pressure from rain and snowmelt by at least 10. “The 1963 slide occurred because of the combined effects of a rising reservoir and increases in piezometric levels as a result of rainfall.”
Why so fast
“The high velocity acquired by this slide mass was easily the most unforeseen phenomenon associated with the Vaiont Slide.” Table 12 gathers the estimates: 17 metres a second from the distance travelled; 25 to 30 on a rheological basis, from Pacher, reported by Müller; 32 from the wave height; 25 from an estimated twenty-second duration; 50 as an upper bound from the height to which the rock at the toe was lifted; 20 to 30 from external dynamics; and Anderson’s own, 20 to 25, from the distance travelled with friction and pore pressures. “It appears from this table that the total displacement, approximately 400 m horizontally, took place in less than a minute and that the maximum velocity of the sliding mass was 20 to 30 m/sec.” Anderson’s reconstruction of the first eighteen seconds has the mass at 0.38 metres a second after ten, 1.6 after twelve, 6 after sixteen and 10.8 after eighteen, by which time it had gone 35 metres and its factor of safety was near zero.
For the slide to travel so far and so fast, about half its resistance had to disappear once it began to move. The clay was already at residual strength and could not lose more friction; what could change was the water pressure in it. Anderson’s mechanism, worked in Appendix F, is heat. Sliding on a thin layer of clay generates heat along the plane; the water in the clay expands and, because clay is nearly impermeable, cannot drain; its pressure rises, the effective stress on the plane falls, the mass accelerates, and the faster it goes the more heat it makes. “Increased fluid pressures due to friction-generated thermal effects are slower to dissipate in low permeability materials. Hence, slides resting on clays, such as those present at Vaiont, are more susceptible to this strength loss mechanism than those resting on more permeable and thinner strata.” A slope on gravel would not do it. A slope on clay could, and did.
What they concluded
The conclusions are short. “Casual studies of important precedent case histories, such as the Vaiont Slide, should not be accepted by the geological and geotechnical professions.” Any dam-site investigation “should include a detailed study of the proposed reservoir slopes”, and old slides found there should be evaluated under reservoir conditions. “The lesson afforded by Vaiont need not be relearned by another generation.” The slide was an “Act of God”, Müller’s phrase of 1963, “only in-so-far as the frequency and duration of the precipitation was concerned”; its history and collapse “can be examined in quantitative terms”, and conventional methods work “if the input data are consistent with the geologic and hydrologic controls”.
“The greatest gaps in the data accumulated on the Vaiont Slide involve: (a) the lack of substantive water pressure data, and (b) the lack of reliable movement records along the failure plane. Fluid pressure measurements taken from piezometers installed at multiple levels within and below the slide would have provided the essential data for correlation with slide movements and reservoir levels. Reliable measurements of the depth of the failure plane and the magnitude of displacements along it would have helped to confirm the depth and size of the slide mass.” And: “In hindsight, it appears significant that an early (1960-61) diagnosis of the kinematics of the lower part of the slide as similar to that of some glaciers (having zero horizontal velocity at the base increasing to a maximum at the glacier’s surface) led those involved to divert their attention away from field exploration required to locate the failure plane, as well as away from instrumentation and analytical efforts. Subsurface borehole deformation measurements would have shown the error in this hypothesis.” The last sentence of the report is one line. “The accumulated evidence also suggests that the slide could have been stabilized by drainage.”
Sources
- A. J. Hendron Jr. and F. D. Patton, The Vaiont Slide, a Geotechnical Analysis Based on New Geologic Observations of the Failure Surface, Technical Report GL-85-5, US Army Corps of Engineers, Waterways Experiment Station, Vicksburg, June 1985. Volume I, Parts I to X, Tables 7, 8, 12 and 13, and Figures 29 to 33; Volume II, Appendices E, F and G. US government works, read on archive.org. Volume I, Volume II