Dangers at embankment dam boundaries and embedments3 October 2008
Embankment dams may be damaged at structural and foundation interfaces and where fill is penetrated by conduits, instrumentation, or other potentially disruptive features. Robert B. Jansen explains how problems have been caused by displacement and deterioration of hidden and incompatible elements
Significant numbers of embankment dams have failed from either external or internal attack. The threat of inevitable floods and earthquakes, which give few useful signs of their approach, is different from that of the already-present but unseen – and often more immediately dangerous – flaws within a dam and its foundation.
Embankments differ from other dams in their composition of deformable natural materials that are largely inaccessible once embodied in the fill. Behaviour of earth and rockfill zones will be more predictable if they are free of penetrations or other encumbrances. Facilities located upon, against, or inside the dam body add to uncertainties and thereby to the possible modes of failure. Dam engineering requires reducing unknowns to a minimum.
The long history of earthfills and rockfills shows that many of their problems were spawned by designers through inadequate safeguarding of embankment boundaries and introduction of unreliable components into the hearts of the dams. Vulnerability may be found at internal zone limits, foundation contacts, structural faces, or works buried in the fill. An embankment should be zoned for internal stability and capable of conforming naturally to its site without appurtenances or foundation characteristics that might hamper its behaviour.
Structural features integrated with embankments
Damaging settlement and cracking has occurred at concrete facings on rockfill dams built before the 1970s by dumping and sluicing in high lifts. More recent rockfills compacted by vibratory rollers and given improved zoning and face slab and plinth details have performed much better. Methods for repair of faces on the early dams have evolved from concrete patching to filling of cracks with expansive mortar, mastic, or clay-bentonite slurry to rubber, shotcrete, or synthetic membrane overlays. Advanced technology is exemplified by the work completed in 2005 on the 328ft high (100m) Salt Springs Dam in California, a dumped fill built in 1928-31. The extensively cracked concrete face was covered with a flexible PVC membrane system consisting of a geocomposite placed over a thick geotextile layer. This anchored overlay is capable of accommodating continuing deformation of the embankment while reducing leakage to an allowable limit.
Some embankments built in the 20th Century as replacements for old dams incorporate parts of the earlier works. The design of California’s 490ft high (149m) New Exchequer Dam, a concrete-faced rockfill constructed in 1964-66, used its predecessor, a 326-ft-high (99m) gravity dam, as the upstream toe block. An ineffective flexible asphalt-impregnated joint seal was provided where the slab met the downstream face of the old dam 185ft (56m) above its base. Settlement of the rockfill caused severe separation and spalling of slab joints and leakage reaching a maximum of about 490cfs (14m3/sec) in 1967. Repairs in that year were concentrated on sealing the opened joint between the two dams by underwater placement of bentonite-enriched earthfill. This was effective with later replenishment of lost material and addition of a geotextile blanket reducing total leakage by 99%. However, as the dam continued to settle, other face joints needed remedial work, which has included concrete filling, new flexible waterstops, and membrane covering.
Outardes 2 Dam in Quebec, a 180ft high (55m) concrete-faced rockfill built in 1976-78, also has an old concrete dam as a toe block. Benefiting fully from the important advances in rockfill engineering in the intervening 12 years after construction of New Exchequer Dam, this Canadian dam has given superior performance, with negligible deformation.
Parapets on embankments must be designed to withstand waves, as well as water storage pressures if the reservoir encroaches on freeboard. Reinforced concrete walls, including those with concrete buttresses and/or earth bolsters, have served this purpose. To protect against erosion by accidental spill over such a parapet, paving of the top of the fill might be advisable. The design of any parapet requires focus on potential environmental effects, including amplification of seismic forces and weathering from seasonal temperature variations, as well as embankment settlement and erosion. To be a reliable part of the reservoir barrier, the parapet must be an integral continuation of the impervious element of the dam. To ensure this continuity, the embankment zone which serves as its foundation has to be strong, tight, and well protected from erosion or other degradation. Use of a wave wall for sustained storage is usually a departure from original plans. In most cases, the freeboard area of an embankment remains above the water level for long periods of time. Exposed earthfill therefore may crack from desiccation and may be weathered and riddled by rodent tunnels. Close inspection and remediation may be needed to make it capable of supporting a parapet.
The inherent discord of dissimilar structures has led to many weaknesses. Some spillway and outlet towers at embankments have interacted poorly with surrounding fill. The 1916 collapse of one-year-old Bila Desna Dam in the former country of Czechoslovakia involved incompatible deformations at the intersection of the outlet gate tower and the earthfill’s core, resulting in destructive leakage.
The 106ft high (32m) gate tower at Baldwin Hills Dam in California, founded on a weak sedimentary formation, passed through the precarious layered lining and drain systems covering the reservoir interior. Concrete channels leading to the gates were constructed as notches through the lining with walls and floors abutting the tower without seals. These channels, placed on compacted earthfill, settled when the reservoir filling started. To reduce leakage, rubber gaskets were installed in the channel joints with the tower and an attempt was made to grout the lining drain in that area. The tower was near foundation faults opened by subsidence. In the 1963 reservoir break, the tower had undergone minor tilting and vertical movement, but remained erect as the waters escaped through the earth dam in a gash aligned with the faults. While it survived as other reservoir features failed, the details of the tower and its connecting conduits showed vulnerability where they crossed the fragile lining. The breaching involved fault disturbance from general ground movement and not earthquake. However, the region has been seismically active. At another time, with earthquake, the tower’s intersection with the lining might have been a contributor to failure.
In the 1971 quake that caused liquefaction and sliding of the upstream slope of California’s Lower San Fernando Dam, one of the two freestanding concrete outlet towers broke near its base and toppled into the reservoir. That tower was in the area of the moving earth embankment. The other tower, away from the slide, did not fall.
Rouchain Dam, a 197ft high (60m) concrete-faced rockfill built in 1974-76 in central France, has a vertical spillway tower rising to full height through the dam near its upstream toe and connected by a bend to a conduit along the base of the embankment. In the 1980s, continuing leakage that had begun with the 1976 reservoir filling necessitated remedial work on the slab and its joint with the tower. The facing joints were repaired and then the slab was covered with an impregnated nonwoven polyester membrane. The critical joint around the tower, which had displaced several inches so that waterstops were torn, was improved by a flexible membrane seal designed to accommodate differential movement of the dam and the tower. After the modifications in 1983 and further sealing underwater of the tower joint in 1984, leakage was at an acceptable level.
Spilling over embankments
When spillways for embankment dams are inadequate, safe and economical remedial alternatives may be limited. If the present facilities cannot be enlarged, the additional capacity preferably should be developed by works located away from the dam. At some projects, however, embankments have been armoured for overtopping. This previously objectionable practice has won some endorsement for infrequent discharges over lesser dams, partly because of the increased quality of roller compacted concrete. The safety of overlays for higher embankments remains unproven. Caution is required in evaluating the possible effects of high velocities, overflowing debris, settlement, cracks, underseepage and uplift, and sliding on slopes. Armour must have underdrains that are kept clean by effective filters. Protection must also be provided against erosion of the dam at the overlay edges and in the approach area of the embankment slope immediately upstream. Safe dissipation of the discharge energy at the toe of the dam would be imperative if this kind of design is pursued for large structures.
Structures abutting an embankment, such as spillways, diversion works, power intakes, and fish ladders, should have surfaces designed so that the fill can be compacted safely and effectively by heavy equipment. Sharp or reentrant corners must be minimised. Walls providing lateral support for fill must be designed to limit deflection and leakage.
Leaks from adjoining hydraulic conveyances have also eroded fills. Timely awareness of the disorder may require meticulous monitoring. At the Walter Bouldin power project in Alabama, the 175ft high (53m) concrete intake is joined at each side by earth dikes that wrap around the downstream corners of the structure and meet over the concrete-encased penstocks to complete the continuous outer face of the reservoir embankment. At this facility one night in November 1993, an embankment drainpipe and a powerhouse drain were found to be discharging dirty water. Muddy flow was seen in the tailrace. Piezometric levels in the fill over the penstocks were rising rapidly. In an immediate round-the-clock response, investigation was made of possible leakage paths at the intake, the penstocks, and the interfaces of the intake with the embankment. Inspection of the penstocks and testing of fill adjoining the ends of the structure disclosed no significant deficiencies. Vertical holes drilled into intake joints intercepted escaping water. Dye tracing showed flows between the holes and the embankment seepage weir and the powerhouse drain that were carrying eroded material. Chemical grouting of the intake joints reduced weir and piezometer readings to levels lower than historical ranges, indicating effective solution of a problem caused by long-term leaking through contraction joints between intake monoliths.
Many embankment dams exist with internal conduits installed in a time when their drawbacks were unknown or lightly regarded. Failures caused by old outlets with built-in flaws have been common. In some cases, efforts to save the facility were unsuccessful because of the lack of upstream gates or valves for depressurizing the conduit. Engineers today are confronted with the need to rehabilitate such works that still survive. The poor record of buried pipes weighs in favour of placing outlets away from embankments. A safer alternative is a tunnel in an abutment. Where this is unfeasible, effort must be made to minimise effects of conduit malfunction. An outlet or spillway conduit passing through or directly under an embankment might pose the greatest project risk.
A buried conduit must be able to adjust to the altered profile resulting from foundation and dam deformation. It must be designed to compensate for differences in embankment settlement between the toes and the crest section. In the downstream reach of an outlet conduit where internal pressure exceeds external hydrostatic pressure, leakage could pass into the embankment through joints or cracks. In such cases, well-located control gates may be essential safeguards.
Untrenched outlet pipes located under embankments have required special care in selecting and preparing foundation. In some projects, continuous concrete bedding has been used to spread the conduit load. The old practice of supporting pipes on piers or collars has been abandoned. Concentrated loading at the supports and the open spans between supports resulted in sagging pipe and voids in backfill.
In the past, the conventional design for controlling seepage along buried outlet pipes featured concrete cutoff collars. These made backfilling difficult and were conducive to differential settlement. The favoured design now has a filter around the pipe rather than a rigid collar. Even though the use of filters in place of cutoff collars has improved seepage control, any soil cover might erode. Inadequate compaction of conduit backfill has resulted in backward piping, which occurs where water passes through earthfill to discharge under a steep gradient. Seepage pressures at the point of emergence initiate erosional tunnelling proceeding upstream to an inlet that then becomes increasingly enlarged.
The inherent weaknesses of certain kinds of conduits are well known. Although some of these have benefited from recent improvements in design and manufacturing and are used in sewers, culverts, and other kinds of low-head service, they are not favoured for the severe conditions of dam projects.
Pipes made of cast iron, clay tile, or asbestos-cement, some of which are still found at ageing dams, have fractured under embankment and equipment loading. Clay tile was used often for toe drains, laid with open joints for collection of seepage but which also permitted entry of foundation and embankment materials.
In old embankments, wide use was made of buried corrugated metal pipe (CMP) for outlets and drains. However, its shortcomings, often attributed to poor workmanship, have led to limitation of its application in new or rehabilitated dam projects. CMP is susceptible to buckling and joint separation. A principal failing in adverse environments is rapid corrosion, accompanied by disbonding of linings and coatings.
Welded steel pipe has been used in outlet facilities. Corrosion was a common defect in old projects. Buried steel, even if coated and lined, may be damaged by electrolysis. However, the life of steel outlet conduit in embankments has been extended by measures such as coating, lining, cathodic protection, and concrete encasement. At sites where sulfates or acids are present, special cement mixes provide resistance to attack on concrete and mortar.
Unencased precast concrete pipes have disadvantages when placed in or under embankments. They may not be suitable under high loads or in aggressive environments unless given special protection. Problems include cracking and joint offset and separation. Joint gaskets can open as the fill deforms. Corrosion of concrete conduit reinforcement can cause spalling.
Cast-in-place concrete has served adequately in buried conduits at many dams. Some others have undergone cracking and deterioration, typically due to underdesign, poor workmanship, low-quality concrete, and/or attack by deleterious chemicals. For a new project, durable liners and dense concrete with special cements and protective admixtures enhance performance. Measures to control cracking of cast-in-place conduit have included limiting the distance between joints, continuous reinforcement through the joints, and placing in alternate sections to allow time for shrinkage before successive concrete placements. Steel liners have been installed with joints left open until embankment construction has been completed and welding can be done under more stable conditions. A lined outlet in a carefully excavated trench in sound rock on an alignment without sliding potential could be an acceptable alternative to a tunnel in an abutment. Vibrated concrete fully enclosing the conduit and extended to the rock surface without unconforming protuberances effectively buffers the embankment from the conveyance. Cages of heavy steel bars in the encasement reduce vulnerability to cracking under the weight of the dam and the reservoir. Waterstops and continuous longitudinal reinforcement can control joint separation and leakage. A conduit thus embedded in rock is better protected from damage by construction equipment and eliminates the soil backfill erosion problem that was common in old projects.
Plastic pipe is used increasingly in new outlets and drains. For outlets in significant projects, it is commonly encased in reinforced cast-in-place concrete. High-density polyethylene (HDPE) is often favoured for these works and for sliplining old outlet pipes. In water projects, temperature ranges are well within HDPE’s allowable limits, which might be problematic in some other utility services. Its heat-fused joints are typically superior to the bell-and-spigot gasketed joints of some polyvinyl chloride (PVC) pipe.
Even with multiple defences, conduit security still depends on the quality of its foundation. At the 180ft high (55m) Blue Ridge Dam in Georgia, US, a semihydraulic earthfill completed in 1930, a problem developed at a 14ft diameter (4.27m) concrete-encased riveted steel penstock under the embankment. Where it crossed a low-strength foundation zone, the penstock settled differentially. During unwatering, it bulged from external water pressure, separating from the encasement. Penstock repair was made by installing an interior girder. Further work has been under way to reduce dam leakage generally and at the penstock and to improve other project facilities, including an additional spillway in the 1990s and a tunnel outlet through an abutment in 2004.
Movement of a buried conduit may be substantial as it accepts the loading of the dam. In 1969 during construction of Meeks Cabin Dam, a 184ft high (56m) earthfill in Wyoming, foundation settlement and spreading caused concrete outlet conduit joint openings as large as nine inches (23cm), and an aggregate joint separation of 3.3ft (1m). Steel reinforcement and rubber waterstops were pulled loose. Steel liner joints were damaged. Some further foundation spreading, attributed to high pore pressures in weathered shale underlying glacial till at the site, occurred about a year later soon after embankment construction had been completed. This outlet, located along the underside of the embankment, includes a steel-lined concrete conduit between the intake structure and the emergency gate chamber, downstream from which is a concrete horseshoe-shaped conduit housing accessible steel outlet pipes. Rehabilitation of the outlet and the embankment enabled project completion in 1971.
Long-time convention, recently reexamined, has provided for placement of instrumentation inside embankments during construction. This can create potential for leakage paths along instrument tubing and cables in horizontal trenches and/or in vertical risers, as well as along benchmark columns. The drawbacks of vertical installations with lightly compacted backfill have been demonstrated at many projects, including Tarbela Dam in Pakistan, W.A.C. Bennett Dam in British Columbia, and dams in western Quebec. Their backfills settled, creating sinkholes centered in low-density areas around instrument columns. Under such conditions, damage could encroach on filters adjoining a disturbed core. The effective width of the impervious zone could decrease seriously. In the upper part of the embankment where zones are thin, deterioration of the core might open short seepage paths and leave retention of the reservoir to the weakened filter and the downstream buttressing zones. The degree of damage could be hidden by arching of the fill until collapse. The column of degraded material could extend deeply to permeable foundation or to coarse embankment layers, allowing leakage to find the most inviting channels. Dams are better served without instrumentation that introduces such weaknesses.
Projections into embankments
Cutoff walls and galleries, as well as conduit concrete encasement not finished flush or shaped to conform with the foundation contact, may act as obstacles to an embankment’s adjustment to its site. Restriction of movement could damage the fill and cause shearing and/or cracking of protruding features, which may also result from passage of heavy construction equipment. Concrete structures in embankments may be detrimental because of arching and consequent hydraulic fracture of adjacent fill or leakage along contacts with loosely compacted backfill.
A notable example of structural cracking occurred in the mid-1960s during construction of the 770ft high (235m) Oroville Dam, an earthfill with an inclined clay core based on a 900ft long (274m) concrete block which extends above the rock foundation to a height of 120ft (36.6m), including a 50ft high (15.24m) unreinforced concrete parapet with a 1.2:1 sloping downstream face. The thrust of the high fill on that projecting face caused rotation about the heel of the block and longitudinal cracking with openings of several inches, tending to sever the parapet from the rest of the block. To control reservoir leakage into the galleries and through the block, its cracks and monolith joints were sealed by injection of cement grout. Instrumentation showed this remedy to be successful.
The various kinds of thin walls or diaphragms placed as internal barriers in embankment dams and in their foundations have generally provided effective seepage control. However, the benefit of the differential water pressure levels that they establish is accompanied by potential damage at their edges and at leaky defects where steep hydraulic gradients may be conducive to erosion of adjoining materials. Imperfections at the foundation and other contacts and in the barrier itself may be worsened by the hydraulic surcharge and consequent erosive leakage. Wall deformation and cracking due to embankment adjustment and/or seismic load could increase the uncertainties.
Steel sheetpile cutoffs are still found in old dams. Unsatisfactory performance has made such seepage barriers unacceptable for permanent installation in major projects. Corrosion is a principal problem. Piling driven to even shallow depths into firm foundation may deviate from interlock. Sheetpile projecting into an embankment may be harmful if it is perforated through deterioration or is loosely joined so that leaks erode fill on its downstream side. Cutoffs that do not extend longitudinally to full closure can allow seepage to flow around their ends under elevated head and thus create a potential for concentrated erosion.
Flexible synthetic sheets have been used increasingly in embankments for seepage control. In dams typically intended to last hundreds of years, the durability and service life of these embedded manufactured products are of principal concern. Survival depends on the materials, the manner of placement, the function of the installation (e.g., zone divider, filter, drain, reinforcement), and the effects of embankment deformation and cracking. Geosynthetics implanted in an earthfill become integral parts of the dam, usually without access for inspection or ready repair. They must be resistant to the various forces and processes that cause their deterioration. Clogging of filters and drains may reduce their effectiveness to intolerable levels.
Internal geosynthetics constitute discontinuities that may have lower shearing resistance than the neighboring zones of natural materials. Some multilayered systems may undergo differential slipping of the sheets. Synthetics can fail if their junctions with the dam foundation are deficient. Hydraulic gradients at that contact are high, so leakage could escape at damaging velocities.
Use of synthetics as permanent members in an embankment’s zoning has been under way for little more than thirty years and, while much has been learned, the relatively short experience has not resolved all questions as to dependability and durability. Embedment of geotextiles as internal filters in big dams is not generally accepted as a satisfactory alternative to well-designed composites of hard sands and gravels, which have been thoroughly proven in long-time service.
Synthetics have also been placed on the upstream slopes of embankments, and as liners for reservoirs (e.g., San Joaquin Reservoir, California, 2004). Such elements can be accessible for inspection and repair, during drawdown or by divers, which is an advantage not shared by deeply buried synthetics. Access is also a merit of lightly covered geotextiles in drainage works.
Embankment drains installed in early projects, often of crude design and workmanship, characteristically have suffered loss of capacity through deterioration, clogging, and settlement. Collector pipes for some drain systems were extended so far into and under the fill that they were inaccessible for observation or remedial work.
At Baldwin Hills Dam, the rigid clay tile pipe drains and concrete drainage inspection chamber under the reservoir lining were dangerous features in a site undergoing severe ground subsidence. The design of the lining with layers of asphaltic pavement, compacted earth, porous concrete, and asphalt membrane was inappropriate for a formation of loose and erodible silts, sands, and gravels. In the morning of the failure in 1963, there was muddy, increasing flow from the inspection chamber through a blowoff pipe discharging into the spillway conduit in the right abutment of the dam. A few hours later, the reservoir was lost and the still-embedded upper section of the severed spillway was visible in the right face of the breach.
At an embankment’s foundation contact, especially under the core and transitions, extra construction effort is often required in excavation and concrete filling to ensure sound rock without vertical or overhanging surfaces or open cracks. Abutments must be shaped to facilitate access by heavy compactors and to retain earthfill so that interfaces remain tight.
The 1976 failure of the 405ft high (123m) Teton Dam in Idaho was a classic demonstration of the need for such painstaking foundation preparation. The volcanic rock penetrated by the steep key trench in the breach area had wide unsealed joints. Projections and recesses in the trench sides prevented thorough compaction of its erodible earth backfill. The unprotected foundation contact, with its open joints as well as loose rock debris left in situ, passed heavy leakage that carried away vital materials and led to quick collapse of the dam.
The bedrock at the 100ft high (30m) Logan Martin Dam in Alabama, a composite of concrete and earthfill sections completed in 1964, is fractured dolomite with numerous cavities from solution. Some erosion resistance at the earthfill’s contact with this formation in the river channel was provided by a reinforced concrete slab in an area on the line of the grout curtain. Leakage through the unsealed remainder of the foundation rock contact caused erosion and sinkhole activity in the embankment. Remedial grouting over many years, intensified in 1991, has given the underside of the dam substantially increased protection.
Embankments not founded on rock may pose challenging problems. Meeks Cabin Dam in Wyoming, 184ft high (56m) and about 3100ft (945m) long, was placed on glacial till containing permeable sand, gravel, and cobbles. The till is underlain at varying depth by weak shale whose instability resulted in dam and outlet damage during the construction completed in 1971. For more than 20 years, seepage, slope sliding, and upstream sinkholes in the left abutment required continuing work to protect against piping. Major improvements were accomplished in the mid-1990s, the principal feature being a deep plastic concrete cutoff wall 825ft (251m) long through the embankment and the permeable foundation elements in that area. As a result, site conditions have stabilized.
Robert B. Jansen, consulting civil engineer, was chairman of USCOLD (now US Society on Dams) and director of design and construction for the USBR and the California Department of Water Resources. A member of the National Academy of Engineering, he has authored and edited books on dams, including Advanced Dam Engineering and Dams and Public Safety.