The collapse of the Sella Zerbino gravity dam

Abstract

When a severe flood wave completely filled the Ortiglieto reservoir on August 13, 1935, the 14 m high “Sella Zerbino” secondary dam failed catastrophically causing > 100 casualties. Both of the dams, Sella Zerbino-Zerbino Saddle and Bric Zerbino-Zerbino Peak (Fig. 1) were overtopped but only the Sella Zerbino failed whereas the main barrage did not suffer any damage. The lawsuit that followed this tragic event ended with a full acquittal of the dam's designers since the plaintiff experts succeeded in demonstrating that the collapse was due to an extreme rainfall storm of unpredictable intensity. The case was then officially closed and still today the failure of the Sella Zerbino dam is attributed to the unpredictable hydrological event. Recently, Natale and Petaccia (2013) re-examined the case assessing the capacity of the flood spillways which equipped the Bric Zerbino dam. This paper thoroughly reviews the mechanics of the collapse of the Sella Zerbino dam focusing on the stability of the structure. The water pressure underneath the dam and the poor quality of the foundation rock is believed to have played a major role in the sequence of events that ended in the collapse of the barrage.

1. Introduction

In Italy, the construction of large dams started at the end of XIX Century when the second industrial revolution was just beginning (ANIDEL, 1961). The Gleno disaster, occurred in 1923, is considered as the first dam collapse happened in Italy. It is believed to have been caused by foundation instability combined with changes in the construction methodology. It provoked > 400 causalities (Pilotti et al., 2011).

According to ICOLD (1974), the main causes of failure of gravity dams are foundations deficiencies and inadequacy to release flood through spillways and outlet works. Poorly-designed spillways are often causing dams failures (ICOLD, 1991).

The Spanish Puentes dam was a 286 m long and 50 m high concrete dam, laying on wooden pillars inserted in a sandy ground. In 1802 the dam suddenly fell at the first filling of the reservoir and killed 608 people.

The Bouzey dam (France) was a dam 525 m long and 22.7 m high that created a reservoir of 7 million m³ (Smith, 1994). A 5 m deep cutoff wall was built to improve the rather leaking bedrock. In 1895, at the initial filling of the reservoir, water started to spout out from the bedrock and the dam slipped forward with a maximum displacement of 35 cm. The dam failure killed 85 people.

The St Francis dam (Begnudelli and Sanders, 2007) was a curved concrete gravity dam (California, USA). The dam was 57 m high, 213 m long, and the reservoir stored 47 million m³. In 1928 when the reservoir was first filled to its crest the dam failed because of the poor quality of the bedrock: the westward abutment was built on a fault while the eastward one was built on mica schist interspersed with talc.

The Sella Zerbino dam in Northern Italy is considered a classical example of poorly designed spillways because the 1935 peak flood discharge was 3 times bigger than the design discharge.

The design of the Bric and Sella Zerbino dams was completed in 1926 under the “Dam Construction Rule No. 1309” emitted by the Ministry of Public Works April 2, 1921, already in force. This Rule required that the water uplift to be considered in the dam stability analysis.

After the Gleno dam failure in 1923, the Italian Government appointed a Technical Commission (December 6, 1923) to carry out a detailed investigation on the safety of existing dams to determine the need for possible retrofitting measures. On December 31, 1925 the Technical Commission submitted to the Government the new Dam Design and Construction Standards which were promptly approved. Only in November 1st 1959, these technical regulations were superseded by the Decree n. 1363; new Construction Standards were issued in March 24, 1982 to update the previous ones. In Italy, the design of new dams and the assessment of existing ones is ruled by the Ministry Decree published on August 7, 2014. The major innovation with respect to the previous technical regulations is the shift from a prescriptive design philosophy to a performance-based design approach. The lessons learned from the catastrophic events occurred over the past 150 years were somehow poured into the new regulations. An important aspect that is now fully recognized is the need to properly recognize from the beginning of the design procedure that hydrological, hydraulic, geological, geotechnical and seismic processes are strictly connected.

2. Design history of Sella and Bric Zerbino dams

The Ortiglieto reservoir stored the waters of Orba basin which covers an area of 142 km² on the leeward side of the Liguria Apennine in Northern Italy (Fig. 2). The design of the Molare hydropower plant changed several times in the 28 years period from the initial design to the start up of the plant. The first project to exploit a discharge of 0.35 m³/s from two Orba River tributaries, dates back to 1899 (Zunini, 1899). In the following sections, the design of the main and secondary dams is discussed separately.

Fig. 2.

Ortiglieto reservoir, location of Sella and Bric Zerbino dams.

2.1. Bric Zerbino main dam

The first project of Bric Zerbino dam, dating 1899, aimed to create a reservoir with storage capacity of 8.1 hm³. The normal water elevation was 311.00 m above the sea level (a.s.l.); the maximum water elevation was 313.00 m a.s.l. Floods were discharged by a gated lateral spillway. In July 1912 after the permission to exploit the reservoir for hydroelectric purposes was granted, a call to accrue the reservoir volume up to 12.25 hm³ was put forward. The new project increased the maximum reservoir elevation to 316.00 m a.s.l.

The dam was 40 m high and 50 m long and its flood spilling capacity was 328 m³/s. This project was approved in 1915.

On April 13, 1921 the construction manager filed for increasing the reservoir capacity to 16.15 hm³ with 320.00 m a.s.l. maximum water elevation. The 45.5 m high dam was equipped with Heyn siphons and a bottom outlet for an overall discharge of about 660 m³/s.

In 1923 the Heyn siphons were replaced by 2 groups of 9 broad crested weirs, operating at 322.00 m a.s.l. The discharge released was about 800 m³/s for a reservoir level of 315.00 m a.s.l.

The May 1924 upgrade increased the reservoir volume to 18 hm³ and the maximum water elevation to 323.00 m a.s.l. A battery of 12 Heyn siphons released 500 m³/s (Petaccia and Fenocchi, 2015).

In 1925 the projects of the bottom outlet and the side spillway were presented. The bottom outlet was regulated by a bell valve and could release up to 150 m³/s. The design discharge of the side spillway, located rightward of the dam, was 110 m³/s. A high pressure bottom outlet of 55 m³/s was also present.

The side spillway was then modified in 1926, since a flood event evidenced its insufficiency. The capacity of the modified spillway increased to 160 m³/s.

Finally, on August 13, 1935, the Bric Zerbino dam was 47 m high, that is 40% higher than the original project, and 191 m long (Fig. 3). The distance between the main and the secondary dam was 500 m.

Fig. 3.

Downstream view of the Bric Zerbino dam, 13/8/1935.

2.2. Sella Zerbino secondary dam

The 1899 project of the Ortiglieto plant envisaged a straight low sill 78 m long located at Sella Zerbino to evacuate floods. The releasing capacity of this weir, equipped by 24 couples of gates 1.5 m high, was 400 m³/s.

The 1921 the construction manager proposed to change the weir into an Ambursen non-overflow gravity dam: due to poor quality of the foundation rock spillways were excluded. As the Dam Office of the Ministry rejected this ill-advised proposals, a revised project of a gravity dam was presented on May 21, 1924. No additional geotechnical investigations were accomplished despite the highly deteriorated characteristics of the bedrock. As later discussed, this played a major role in the disaster. The four blocks of the dam were separated by three contraction joints.

On August 13, 1935 the final shape of Sella Zerbino dam was as follows: height 14.5 m, length 109 m, shoulders made as solid walls 3.5 m broad. The slope of the faces of three central blocks of the dam were: 10% upstream and varying from 75% to 55% downstream (Fig. 4). The secondary dam had no spillways.

Fig. 4.

Downstream view of Sella Zerbino dam 13/8/1935.

A soon as the reservoir was filled, water leakages of about 0.06 m³/s were detected, so that grout injections in the bedrock were called for. The grout curtain did not stop the leakage; in fact the final inspection noticed a considerable water spillage for a reservoir level of 321.80 m a.s.l.: a leakage of 0.017 and 0.005 m³/s from the right and left abutment of the dam was reported. Table 1 shows the historical evolution of the Ortiglieto project involving the two barrages.

Table 1.

Historical evolution of Bric Zerbino and Sella Zerbino design projects.

	Measure units	1898	1912	1921	January 1924	May 1924	1925	1926
Maximum w. el.	m a.s.l.	–	–	322	323	323	323	323
Normal w. el.	m a.s.l.	313	316	320	322	322	322	322
Flood spillways capacity	m3/s	400	n. a.	553	1000	n. a.	815	865
Live storage	hm3	9.5		12.25	16.15	16.15	18	18
Bric Zerbino height	m	33	40	45.5	45.5	45.5	47	47
Sella Zerbino height	m	1	10	14	14	14	14.5	14.5

3. The failure of the Sella Zerbino dam

After a long dry period, at 6:15 a.m. of 13th August 1935 an exceptionally severe rainfall storm hit the Orba basin (Natale and Petaccia, 2013). At 7:00 a.m. the rain intensity increased and kept on without interruptions until 3:00 p.m. The rain reached his highest intensity between 7:00 and 8:00 a.m. and between 2:00 p.m. and 3:00 p.m. At 9:30 a.m. the siphon outlets operated to their full capacity (Petaccia and Fenocchi, 2015). At 10:45 a.m. the water began to enter the side spillway and soon afterwards the flood inflow exceeded the maximum capacity of the spillways. After 15 min the bell valve was clogged up with sediments and debris and there was no way to re-open it. At 12:30 a.m. both of the dams were overtopped. At 1:15 p.m. the Sella Zerbino dam abruptly collapsed (Alfieri, 1936). Historical witnesses confirm that the collapse started from the left side of the dam. The remaining parts of the dam failed sequentially. On August 15, 1935 a new flood wave swept down the wreckage of the left abutment. The dam failure flooded an area of almost 70 km² and caused 111 casualties, 97 of which in Ovada town (Natale et al., 2008, Petaccia et al., (in press)).

4. Geologic and geotechnical site characterization

The geological reports prepared for the design of the two Zerbino dams were mainly referred to the location of the Bric Zerbino dam (Lelli, 1937, Peretti, 1937). Recent surveys on Sella Zerbino site (Capponi, 2014, Bonaria and Tosatti, 2013) highlighted the extremely poor quality of the bedrock. Fig. 4 shows the transversal cross section at the dam site.

The mechanical characteristics of the bedrock were retrieved from the results of the experimental investigation campaign funded in 1980 by Piedmont Regional Administration to assess the budget costs of a multipurpose plant aimed to revitalize the exploitation of the Orba River. The investigation included a series of geophysical refraction surveys (Calvino and Siccardi, 1980) from which a total of 15 subsoil profiles were constructed. The average low-strain elastic moduli of the undisturbed rock formations were then estimated from the measured speed of propagation V_P of compressional waves. The values of V_P typical of a compact rock could only be found below 20 m from the free surface. Three boreholes with continuous sampling were drilled to reconstruct the lithostratigraphic profile along the axis of the Sella Zerbino dam (Fig. 5). The rock formations were distinguished based on the RQD (Rock Quality Designation) geomechanical parameter.

Fig. 5.

Longitudinal vertical cross-section at the site of foundation of the Sella Zerbino dam (view from upstream).

The hydraulic conductivity of the geologic formations was measured through standard in situ Lugeon tests. The subsoil at the foundation site of the dam is locally fractured; therefore a distinction was made between primary and secondary permeability. Furthermore, different mechanical properties of the various lithotypes produce systems of discontinuities with different hydraulic parameters. Fig. 5 shows qualitatively that the bedrock is highly fractured and inadequate to bear the stresses induced by the weight of the Sella Zerbino dam. As mentioned above, this was noted during the construction works when the grout curtain was injected.

The top soil layer is a sandy silt originated from weathering of the underlying bedrock and an highly fractured Serpentinite with talcose silt; its thickness varies from 1 to 4 m below the ground level. A large variability of geomechanical characteristics characterizes the intermediate geologic formation at the right abatement (borehole S3 in Fig. 4). This is also caused by the presence of a mylonite interface in the interior of the riverbed. In the central part of the dam cross-section the intermediate geologic formation has relatively homogeneous properties (borehole S2 in Fig. 4). However, the terminal part of the borehole intercepts the mylonite with some visible Serpentinite pieces after crossing some degraded Serpentinite breccia.

We tested 3 concrete samples of the remains of Sella Zerbino dam recovered during one of our on-site surveys: the results are reported in Table 2. In this table R_ck denotes the average cubic strength, f_ck the average cylindrical strength, f_ctm the average tensile strength and E_cm the average elastic Young modulus of the concrete.

Table 2.

Mechanical characteristics of the cyclopean concrete of the Sella Zerbino dam calculated according to the formulas of the current Italian Building Code (NTC08).

Cubic samples	Rck	fck	fctm	fcm	Ecm
[N/mm2]	[N/mm2]	[N/mm2]	[N/mm2]	[N/mm2]
1	37.20	30.88	2.95	38.88	33,062.50
2	28.87	23.96	2.49	31.96	31,174.51
3	30.17	25.04	2.57	33.04	31,486.59
Average	32.08	26.63	2.67	34.63	31,907.87

5. Possible causes of the collapse of Sella Zerbino dam

Following the collapse of the Sella Zerbino dam, different causes were blamed for the disaster (Accusani, 1936, De Marchi, 1937, Mangiagalli, 1937) including:

• 1.Failure of the internal body of the barrage due to inadequate concrete resistance
• 2.Instability due to the uplift force. The design did not consider this force, even though it was required by the technical Standard of that time (Ministry Rule No. 1309 of April 2, 1921)
• 3.Instability due to the scour caused by the plunging of the water overtopping the dam.

Since both the trial report and our preliminary analyses verified that the cause in 1 did not occur, the following sections thoroughly investigate causes 2 and 3.

6. Seepage analysis

The seepage analysis was carried out to quantify the uplift force acting beneath the Sella Zerbino dam and to assess the overall stability of the barrage. The analysis was performed using SEEP/W (Geo-Slope International Ltd., 2007); this finite element-based code computes the flow net in a two-dimensional porous medium. The spatial domain was extended approximately 40 m upstream and downstream of the dam and was discretized by 1718 quadrangular elements with side ranging from 0.6 to 1.0 m. The hydraulic conductivity and other hydraulic parameters introduced in the SEEP/W model were derived from geotechnical characterization and using empirical correlations (Tong et al., 2014). As an example, Table 3 reports the values adopted for the various geological formations in cross section 5. Fig. 6 shows the finite element grid used in the SEEP/W model.

Table 3.

Hydraulic conductivity in the SEEP/W model for saturated bedrock.

Geomaterials	Conductivity
[m/s]
Mylonite	5 ∗ 10− 4
Talcose silt	1 ∗ 10− 4
Fill from construction site	1 ∗ 10− 3
Sediments	1 ∗ 10− 3
Fractured Serpentinite	1 ∗ 10− 4
Weathered silt with Serpentinous breccia	3 ∗ 10− 4
Fractured Prasinite	1 ∗ 10− 6

Fig. 6.

Finite element model of cross-section S5 used for seepage analysis with SEEP/W.

First of all a steady state seepage analysis was conducted with the water elevation of 323.00 and 326.67 m a.s.l. respectively, corresponding to the maximum storage level in the reservoir and the water level attained when the Sella Zerbino dam collapsed. Fig. 7 shows the flow net beneath the dam obtained for cross section S5. Afterwards, unsteady flow seepage analysis was performed to simulate the evolution of the phenomenon from 10:00 a.m. to 1:20 p.m.

Fig. 7.

Steady state flow net calculated with SEEP/W for cross section S5 and reservoir water elevation of 326.67 m a.s.l.

Fig. 8 shows the comparison for cross section S5 (Fig. 5) of the components of the uplift force calculated under steady state and transient conditions. The results of the two types of analyses are rather close to each other: the storage factor is irrelevant due to the almost large values of the hydraulic conductivity of the bedrock.

Fig. 8.

Comparison of the components of buoyancy forces calculated under steady state and transient conditions for cross section S5 (Fig. 5).

7. Stability analysis of Sella Zerbino dam

The forces acting on the cross section of the gravity dam are shown in Fig. 9: the destabilizing actions are in red color and the stabilizing actions are in green color. The forces are: weight of the barrage W, weight of the soil above the sliding plane W_t, hydrostatic force on the upstream face of the dam S_i, the thrust of the saturated soil S_s, the uplift thrust S_p, and possibly the ice thrust and the seismic action. The contraction joints attenuate the internal forces directed as the axis of the straight gravity dam. With reservoir full, the dam is subjected to tensile stresses at the upstream face and to compression stresses at the downstream face. The stability of a single element should be assessed for sliding, overturning and strength of the material to withstand strains. The geometry of nine cross sections taken from the drawings of the construction project dated February 8, 1924 located as shown in Fig. 5 was carried out.

Fig. 9.

Pressure and forces acting on section S5 when the dam collapsed (reservoir upstream level: 326.67 m a.s.l).

The assessment of stability of the Sella Zerbino dam is performed according to the Standards: DM No. 1309 of April 2, 1921; DM No. 44 of March 24, 1982.

The first one was in force at the time of Bric Zerbino and Sella Zerbino project and construction. The second one refers to the technical regulations for the design of dams that were in force until very recently in Italy. The main difference between the two technical norms is the shape of the underpressures, which is trapezoidal according to the DM 44/82 while is triangular according to the DM 1309/21 (Fig. 9).

The stability verifications are conducted with respect to two scenarios, which are considered the most relevant in our case: A: maximum water level 323.00 m a.s.l; B: water level at the dam collapse 326.67 m. The destabilizing actions increase considerably when the level of the reservoir increases of 3.67 m. The sliding plane of the structure is assumed horizontal, as shown by the dashed red line in Fig. 8, and the portion of geomaterial W_f above the this line is assumed to move along with the bulk of the dam; this is considered reasonable in view of the predominantly horizontal direction of the upstream hydrostatic pressure, and the highly fractured bedrock whose joints pattern is unknown. This hypothesis was also confirmed by the on-site surveys, since the active river bed is now placed at the former elevation of the upstream face of the dam. The uplift thrust S_p,21 was calculated following the technical prescriptions above mentioned.

Table 4, Table 5 show the results of stability analyses at nine cross sections of the dam (Fig. 5) and for the A and B aforementioned water elevation scenarios. The tables also include the compression stresses in the concrete along the cross-sections of the upstream (σ_upstream) and downstream (σ_downstream) parts of the dam for both A and B water elevation scenarios.

Table 4.

Stability analysis of the Sella Zerbino dam for the maximum storage level (323.00 m a.s.l.)

Table 5.

Stability analysis of the Sella Zerbino dam for the reservoir level corresponding to collapse (326.67 m a.s.l).

The numbers in red color denote a critical situation where the stability condition is not satisfied. It should be remarked that the results shown in Table 4, Table 5 were obtained by considering a block of the dam of unit width along the longitudinal direction of the barrage at each of the nine aforementioned cross sections. The sliding coefficient SS is defined as the ratio between the forces acting along the direction of the x axis and those acting along the positive direction of the y axis. To fulfill the verification, the sliding coefficient SS must be lower than 0.75.

The overturning coefficient SR is defined as the ratio between the stabilizing and the destabilizing forces; to fulfill the verification the coefficient must be > 1.5.

Overall, based on the results illustrated in Table 4, the Sella Zerbino dam can be considered stable for the maximum reservoir level assumed in design. The safety margin however is tight and this should not be acceptable given the level of importance of the infrastructure and the catastrophic consequences of a failure.

The compression stress in the concrete do not exceed the admissible ones in any of the cross sections of the dam for both A and B scenarios. Hence, the structural design of the dam was correct since the geometry of the upstream and downstream faces of the dam in all its blocks allow a correct redistribution of the compression stresses even for an extreme loading scenario which was not considered in design. On the contrary, the safety against sliding and overturning fails to be satisfied at several cross sections for B scenario. Furthermore, the most unstable blocks are those placed on the left side of the dam, viewed from upstream, and this using both aforementioned technical Standards: the fact was confirmed by the few witnesses of the event (Audoly, 1939). From Table 4, Table 5 it appears that the safety conditions of the dam rapidly deteriorate along the left side of the barrage. It is remarked once again that the stability assessment discussed above was performed assuming that the stresses act in the cross section planes.

The change of safety margin against sliding verification of the dam from scenario A and B measures the gradual reduction of dam stability with the rise of the water level in the reservoir. Table 6 shows the results of the stability assessment of the Sella Zerbino dam by calculating the uplift force according to one of the following methods: Technical Standards No. 1309 of April 2, 1921; SEEP/W analysis under steady state conditions; SEEP/W analysis under transient conditions.

Table 6.

Sliding safety conditions of the Sella Zerbino dam assessed with different criteria.

SS	Rule no. 1309		Steady-state conditions		Transient conditions
323.00 m	326.67 m	323.00 m	326.67 m	323.00 m	326.67 m
Cross section S3	0.49	0.98	0.44	0.87	0.47	0.88
Cross section S4	0.47	0.91	0.52	1.04	0.49	1.08
Cross section S5	0.53	0.96	0.51	0.90	0.52	0.89
Cross section S6	0.77	1.33	0.60	1.09	0.59	1.04
Cross section S7	0.55	1.09	0.53	1.07	0.53	1.03

The three criteria are comparable in results. The stability assessment under transient conditions allowed calculating the time evolution of the sliding safety factor. Fig. 10 shows the predicted evolution of failures along the longitudinal axis of the dam. Cross sections S6 and S7, belonging to the second block of the dam, appear to be the most critical.

Fig. 10.

Cross sections sequence of failures along the longitudinal axis of the Sella Zerbino dam.

Let us consider the stability of the blocks of the dam separated by the joints at cross sections S3, S5 and S7 as shown in Fig. 11. The actions taken into account for the equilibrium are the hydrostatic pressure, the uplift force and the weight of the concrete block.

Fig. 11.

Axonometric view of the Sella Zerbino dam with highlighted the cross sections of known geometry along the axis of the barrage (Gianfranceschi, 1926).

The Cartesian reference system used for the stability calculations of the dam blocks is also indicated in Fig. 10. The total uplift force was calculated by integrating the pressures along the foundation.

For the water level at the collapse of the Sella Zerbino dam the block in the most critical stability conditions is contained by the joints 5–7. This confirms that the first block to fail, is the one adjacent one to the left abutment (see Table 7). Again, this is due to the highly fractured bedrock characterizing that part of the barrage.

Table 7.

Stability analyses of the blocks of the dam with the sliding coefficients.

		ΣFo	ΣFv	SS	SSblock
		[kN]	[kN]
Block 4	From cross section S1 to section S2	3817.40	6403.78	0.60	0.87
From cross section S2 to section S3 (joint)	13,692.15	13,663.66	1.00
Block 3	From cross section S3 (joint) to section S4	16,845.37	13,875.63	1.21	1.18
From cross section S4 to section S5 (joint)	11,842.08	10,442.20	1.13
Block 2	From cross section S5 (joint) to section S6	17,102.37	13,934.99	1.23	1.23
From cross section S6 to section (joint) S7	10,359.39	8346.63	1.24
Block 1	From cross section S7(joint) to section S8	19,118.15	17,889.22	1.07	1.01
From cross section S8 to section S9	5023.98	6014.68	0.84

Once open, the breach enlarged toward the block of the right abutment of the dam. On the left bank, only the extreme portion of the abutment was left standing, as evidenced by some old photographs.

8. Bearing capacity of rock foundation

A bearing capacity calculation was performed to assess possible exceedance of the allowable pressure at the contact between the foundation of the Sella Zerbino dam and the underlying geological formation. The computation has been carried out using the procedure discussed in Annex G of Eurocode 7 Part 1 (EN 1997-1:2004/AC, 2009). This is based on a presumed allowable bearing pressure estimated from a qualitative description of rock quality. Table 8 shows the results of the analysis conducted for the water elevation of 323.00 and 326.67 m a.s.l. respectively. The rock was assumed characterized by medium spaced discontinuities with an allowable bearing pressure of 1750 kPa. According to the two technical norms (DM 44/82 and DM 1309/21).

Table 8.

Bearing capacity of the rock foundation along the various blocks of the Sella Zerbino dam.

		qes,323	qes,326
	[kPa]	[kPa]
Block 4	From cross section S1 to section S2	148.51	152.44
From cross section S2 to section S3 (joint)	117.30	122.34
Block 3	From cross section S3 (joint) to section S4	72.94	77.07
From cross section S4 to section S5 (joint)	42.80	43.80
Block 2	From cross section S5 (joint) to section S6	54.21	55.57
From cross section S6 to section (joint) S7	39.46	39.97
Block 1	From cross section S7(joint) to section S8	56.64	57.53
From cross section S8 to section S9	28.91	29.57

The stability of the dam to bearing capacity is assured if the factor of safety (FS), defined as the ratio between the maximum allowable pressure q_lim and the contact pressure acting at the rock-foundation interface q_es, is > 2.5. The calculation give a value of FS equal to 3.12 for a water elevation of 323 m a.s.l and equal to 3.03 for the reservoir elevation that caused the collapse of the dam. Therefore, based on the available information, the rock underneath the foundation had a bearing capacity that it appears adequate to sustain the contact pressure transmitted by the Sella Zerbino dam under both ordinary loading conditions and at the time of collapse.

9. Scour depth estimation

The overtopping of the Sella Zerbino dam was not considered in the design. In the trail, the consultants of the designers identified the scour as a possible cause of the collapse. For this reason, we calculated the maximum scour depth for a water elevation of 326.67 m a.s.l. applying different empirical formula (Liu, 2005) which can be written in the form:

$$\mathrm{t}=k_e\frac{q^x{H}^y}{d^z}$$ (1)

where q is the unit discharge overtopping the secondary dam, d is the characteristic dimension of the average rock, H is the difference between upstream and downstream water surfaces, k_e is a coefficient related to rock resistance to erosion. Table 9 shows the coefficients of the empirical formula (1) and the scour depth computed at cross section S5.

Table 9.

Parameters of the empirical formula and scour depth estimation for cross section S5.

	Parameters
Formula	ke	x	y	z	t(m)
Schoklitsch (1932)	4.75	0.57	0.20	0.32	1.33
Chen (1963)	0.50	0.50	0.25	0.00	1.48
Yu (1963)	0.56	0.75	0.13	0.00	1.48
Yuditskii (1971)	0.51	0.67	0.33	0.33	1.26
Mason and Arumugam (1985)	0.50	0.60	0.15	0.20	0.90

Fig. 12 shows the depth of the scour for cross section S5. This was almost 30% of the refill material in the downstream part, corresponding to 1.2 m.

Fig. 12.

Scour depth on cross section S5 of the secondary dam.

10. Concluding remarks

The Italian dam safety regulations effective on 1935, the year the Sella Zerbino dam collapsed, would avert the disaster if enforced. The several changes to the 1899 design of the Zerbino lead to a final very unsure configuration. The leakages through the bedrock of the secondary dam, detected by the on-site inspections, were dealt with sporadic and ineffective grout injections.

This paper mainly focused on the overall stability of the Sella Zerbino dam first under the conditions assumed in design and then at the time when the barrage collapsed on August 13, 1935. According to the technical Standard, effective when the dams were designed, and more recent regulations, the Sella Zerbino dam was stable to sliding, overturning and exceedance of stresses in the concrete for the design water level only. However, from analyses' results, it appears that the discriminating factor jeopardizing the stability of the secondary barrage is the hydraulic pressure underneath the dam which was ignored by the designer. Overtopping of the Sella Zerbino dam did not seem to affect the stability of the barrage according to the present scour calculations. The sudden increase of the reservoir water level was due to the inadequate capacity of the flood spillways which were not able to release the flood inflow.

A finite element model was applied to simulate seepage beneath the dam as well as to investigate the influence of the fractured rock foundation on the magnitude of the uplift forces and thus on the stability of the barrage. The extremely poor quality of the bedrock seems to have been the cause of the instability of the dam section located at the left side of the barrage (cross section S7) that rapidly brought to the formation of a breach toward the center of the dam. A conclusion that can be drawn from these analyses is that the collapse of the Sella Zerbino dam could perhaps have been prevented whether technical regulations of that time were adopted.

Fig. 1.

Location of Ortiglieto reservoir in Northern Italy.