The case of Cleveland dam

THE Greater Vancouver Water District (GVWD) provides a reliable source of safe, high-quality drinking water to 18 member municipalities and 2 million residents of the Greater Vancouver region, British Columbia, Canada. Water is collected from three mountainous watersheds; Capilano, Seymour and Coquitlam and is delivered by an extensive system of 22 distribution reservoirs, 15 pumping stations and over 500km of supply mains.

Cleveland dam stores approximately 40% of the total regional water supply and is a key element in the system. The dam is located on the Capilano river, approximately 5km north of Burrard Inlet at the northern limit of the Capilano River Regional Park. The area south of the dam is open to the public for recreational activities whereas the watershed catchment north of the dam is a restricted zone. A salmon hatchery is located approximately 300m downstream of the dam, and the downstream area has a large residential population and extensive commercial and industrial development.

Cleveland dam is a 91m high concrete gravity structure set in a deep bedrock canyon. Full reservoir supply level is at el. 146m and the dam crest is at el. 149m. In 1992, the dam was upgraded to pass the Probable Maximum Flood (PMF) and to withstand the Maximum Design Earthquake (MCE) event.

The dam impounds a reservoir extending 4.5km to the north where the Upper Capilano river flows into the reservoir. Typically 0.5km wide, the reservoir surface area is 2.3km2 at full reservoir level. Flow is controlled at the dam by a radial gate spillway, two mid-level intakes and outlets, and two low-level outlets.

The Capilano canyon was considered a potential dam site as early as 1886. A 1940’s site investigation confirmed the feasibility of constructing a water supply dam at the site. The concrete dam was completed in 1954.

Geological setting and seepage history

During the more than 50-year history of Cleveland dam, numerous studies and investigations have better defined the geology and hydrogeology. Dozens of piezometers and other instrumentation have been installed. The complex seepage problem has been effectively managed with the essential input of geotechnical specialists such as Karl Terzaghi, Charlie Ripley and Ralph Peck.

The dam is constructed in a bedrock canyon on the west side of the valley. Immediately east of the canyon is a wide, deep buried valley that extends about 2km (eastward from the dam to the base of Grouse Mountain. The buried valley extends to below sea level and is infilled with a complex stratification of silt, sand, gravel and glacial till. The west canyon bedrock is hard, moderately jointed granodiorite and quartz diorite with numerous inclusions of older sedimentary and volcanic rocks.

The ancestral Capilano valley infill forms the east abutment of the Cleveland dam. These surficial deposits may have been a ‘natural’ dam for earlier water bodies formed at various stages of glacial advance and retreat. The surficial geology is the result of a glacial depositional environment in a near marine or lacustrine setting. Basal infilling includes significant thicknesses of silt, with increases in granular materials as a result of several periods of glacial advance and recession along with the accompanying rise and fall of relative sea level. This process has formed two primary aquifers, the upper and lower aquifers, both susceptible to seepage from the Capilano reservoir. The aquifers are comprised of lenses and layers of sand, sand and gravel, and sand, gravel, cobbles and boulders. These two aquifers are sandwiched between less permeable aquitards that are comprised of silt, till and till-like deposits. Overlying the upper aquifer is a glacial till unit and the Capilano Sediments, a cobbly, bouldery sand and gravel.

The upper aquifer is exposed both on the upstream and downstream slopes of the east abutment within slopes that are typically inclined 35° to 55°. Investigations have shown that the upper aquifer is continuous for hundreds of metres in a radial direction from the east end of the concrete dam. It is near horizontal and transmits 90% of the seepage through the east abutment.

Although the upper aquifer is characterised primarily as a unit of sand, gravel, cobbles and boulders (SGCB), some seepage may be carried by the other permeable strata within the unit. At many locations, this unit contains considerable fine-grained soils and some till-like materials. This wide range of material properties is typical of a very active depositional environment. The upper aquifer is therefore an assemblage of permeable stringers within an overall mass of materials, much of which may transmit only a small portion of the seepage. Of particular relevance to the potential for piping is the presence of a sand layer, averaging 6m thick at the base of the upper aquifer SGCB unit. This sand layer and the overlying very dense material, which may bridge over voids, provides a potential piping pathway through the east abutment.

The potential for seepage, piping and slope instability in the east abutment was recognised at the time of design and construction of the original dam in the 1950s. The original dam therefore included a 1.5m thick till blanket covering the upper aquifer exposure for 200m on the upstream reservoir slope and a single line grout curtain through the overburden to bedrock about 166m into the east abutment. Upon impounding the reservoir, these seepage control measures were found to be inadequate and consequently several drainage adits, several emergency pumping wells and a plastic membrane extension to the till blanket were installed between 1955 and 1967 to help control piezometric levels, seepage and piping in the east abutment. Some of these adits were later backfilled as the seepage they intercepted was minor. In 1991 and 1992, drainage and filtering systems were upgraded in selected adits, existing wells.

Following reservoir filling in 1956, to the late 1990s, at least 120m³ of sand and coarse silt eroded from the east abutment aquifers into the drainage adits. The actual volume of eroded material could be greater, as the sediment traps do not retain the silt and clay fractions, and therefore this quantity has not been measured. The measured soil losses were mainly in lower aquifer drainage adit T470 (30%), and upper aquifer drainage adits T505 (65%) and T506 (5%). Soil losses from T505 and T506 have been negligible since the drain filters were upgraded in 1991 and 1992. Nevertheless, significant open spaces or voids may remain. The volume of these voids may provide adequate storage volume such that internal erosion could continue back from the adits towards the reservoir without further detectable migration of sands out of the abutment. This could result in high piezometric pressures in the downstream slope, which could decrease the stability of the abutment.

Since the late 1980’s increasing piezometric pressures were also observed across the east abutment resulting in an increase in seepage gradients and a 15% increase in flow rates from T505/506 drainage adits. The increase in gradients and flow rates results in an increased risk of piping through the upper aquifer. The purpose of the original seepage control and drainage works was to reduce piezometric heads and gradients across the east abutment thereby maintaining an adequate margin of safety against failure. With the possibility of internal erosion downstream of the original till seepage control blanket, relatively high operating heads have been utilised at the drainage works to reduce gradients and piping potential. As a consequence, piezometric heads in the downstream slope were also greater than desirable although monitoring indicated the downstream slope was stable. Improvements to the seepage control works were recommended in order to prevent migration of piping to the reservoir near the end of the till blanket. Reduction of gradients and piezometric heads in the east abutment were required to improve the factor of safety.

Design and scope of remedial measures

Design criteria
Hydrogeologic modelling indicated that the maximum hydraulic gradient occurred near the upstream end of the original till blanket. Higher local gradients may exist at the end of any seepage pipes extending upstream from the T505/506 drainage adits, but the exact locations of these pipes, if they exist, could not be predicted. The principle design criterion used to determine the length of the blanket extension was reduction of the modelled average seepage gradients at the contact of a ‘high permeability zone’ and ‘low permeability zone’ near the end of the existing blanket by a target of 50% and a minimum of 33% below the gradients in the two dimensional model calibrated to 1999 full pool data. The maximum gradient in this contact zone from the recalibrated model was 8%, therefore, the target average gradient was 4% and the target maximum gradient was 5.5%. Other design criteria regarding blanket extension length required that the blanket cover all known areas of moderate to high seepage potential where high seepage gradients would occur, and blanketing all known open work gravels in contact with the underlying erodible sand stratum.

The calibrated hydrogeologic model indicated that the original till blanket had a conductance (hydraulic conductivity x cross-sectional area / length of the flow path) of about 1×10^-7m/s/m. Blanket extension design was based on a conductance of 1×10^-8m/s/m for the new blanket. The connection between the existing blanket and the blanket extension was designed to provide conductance equal to or less than the conductance of the original blanket.

The blanket needed to be non-erodible and constructible during rainy weather, typical of North Vancouver during the winter/spring construction period. It was required to be stable under full reservoir and drawdown operating conditions and be operational following the MDE.

The reservoir slopes upstream of the original till blanket were very steep, averaging 45° to 55° in the glacially consolidated soils and 35° to 40° in the surficial post-glacial soils. Stability analyses indicated that 37° excavation slopes with 4m wide benches at 10m vertical intervals in the glacially consolidated soils and 35° slopes with 4m wide benches at 10m vertical intervals in the post-glacial soils provided acceptable limit-equilibrium factors of safety for operating, drawdown, and MDE (0.5g) conditions in accordance with Canadian Dam Association (CDA) Guidelines. The temporary construction slopes were designed to meet these criteria and the final overall slope, including benches, was approximately 29°.

Conceptual to final design
Conceptual design studies indicated that the probability of piping failure could be substantially reduced by constructing a positive seepage barrier across the upper aquifer to reduce the seepage gradients. A complete cutoff of the upper aquifer was impractical due to the substantial upstream extent of the aquifer.

The calibrated hydrogeologic model of the abutment identified an area of high permeability near the drainage adits that extended upstream toward the till blanket. The high permeability zone may have resulted from piping of sand into the drainage adits or a natural zone of higher permeability materials. This high permeability zone could be a focal point for further internal erosion toward the reservoir slope. The model also indicated that the existing glacial till blanket was effective in reducing seepage gradients and could be incorporated into any new seepage control measures, but the plastic membrane blanket extension was ineffective and needed replacement. Remedial options considered for reducing seepage gradients included a multi-line grout curtain; a cutoff wall comprised of plastic concrete constructed by slurry trench method, secant piles, jet grouting, or deep soil mixing; an extension of the existing till blanket using earthfill, a concrete membrane or geomembrane; and upgrading the existing observational approach already in place.

The grout curtain option was rejected due to the uncertainty in achieving a continuous seepage barrier. The extension of the observational approach alone was considered untenable, as detection of piping migration with instrumentation was difficult to assure. Therefore a cutoff wall or extension of the seepage control blanket became the preferred remedial measures.

Site investigations found that the sand stratum at the base of the upper aquifer pinched out about 235m upstream of the original till blanket. The hydrogeologic model indicated that a 260m blanket extension was required to limit the hydraulic gradients to the design criterion, (discussed in the following section). Mapping of seepage points on the reservoir slopes during reservoir drawdown and drilling site investigations confirmed that a 260m blanket extension length satisfied the selected design criteria.

The hydrogeologic model was used to evaluate the effectiveness of the cutoff wall and blanket extension options in reducing hydraulic gradients. In the early design stages an earthfill blanket extension was recommended as the preferred remedial solution because of lower capital cost relative to a deep cutoff wall excavated from the existing ground surface or a local excavation, the ability to visually inspect most of the construction, dealing with the problem at its source and the relative ease of future repair work, if required.

The original blanket was constructed of compacted glacial till, however, this material is sensitive to moisture content and significant weather delays were considered likely during the predominantly wet spring period dictated by the construction schedule constraints. Roller compacted concrete (RCC) was selected as the impervious material for the blanket extension as the moisture content could be controlled during mixing and high placement rates were possible. Processed granular materials from the excavation could be used in the RCC thereby reducing the volume of disposal sites and minimising import of materials.

Operational requirements limited the reservoir drawdown to el. 120m, however the upper aquifer extends to a minimum of el. 107m within the blanket extension. Therefore, construction of the RCC blanket extension over the entire vertical extent of the upper aquifer was not possible, and it became necessary to modify the conceptual design to include a cutoff wall to provide a seepage barrier for the lower portion of the aquifer. Excavation of the cutoff wall in panels using a cable clam was determined to be the most economical construction method and plastic concrete backfill was selected in order to achieve the required hydraulic conductance, stiffness comparable to the surrounding glacially compacted soils, and ductility to accommodate small earthquake induced displacements. A bench width of 13.5m was selected as the practical minimum for conventional equipment. The design therefore included a maximum 70m high, benched slope down to the el. 126m construction bench, requiring the excavation of approximately 300,000m³ of material.

The two key components of the seepage control remedial measures are a 306m long vertical plastic-concrete cutoff wall below el. 126m, topped by an inclined RCC blanket extending over the upper aquifer on the reservoir slopes to about el. 138m.

An un-reinforced concrete cap above the cutoff wall provides a transition from the plastic concrete cutoff wall to the RCC blanket. Water stops were provided at the top of the plastic concrete and at the top of the cutoff wall concrete guidewalls, which were left in place.

A short section of compacted glacial till fill ties the existing till blanket to the new RCC blanket extension, and a geosynthetic clay liner (GCL) provided a low permeability barrier at the contact between the top of the RCC and the natural soils.

A rockfill buttress was constructed against the RCC blanket extension to provide ballast support and resistance for stability during reservoir drawdown. About 55,000m³ of rockfill and granular fills were placed over the RCC blanket as ballast and filters.

General construction sequence

The Cleveland dam seepage control project was constructed from March 2001 to July 2002. The overall construction sequence and activities included:

• Construction of a sound-wall barrier along Capilano Mainline Road.

• Excavation of 292,000m³ of soil in stepped benches to create stable slopes and a construction bench at el. 126m for the cutoff wall.

• Application of shotcrete and installation of drains on the excavated slope between el. 138m and el. 126m to provide temporary slope protection.

• Restoration and landscaping of excavated slopes above el. 146m, including installation of a temporary irrigation system to revegetate the upper benches.

• Processing and re-use of suitable excavated material for site fills and RCC aggregate.

• Construction and landscaping of excavation spoil stockpiles within the watershed.

• Development of a quarry within the watershed for aggregate and rockfill.

• Installation of piezometers and monitoring pins along the excavation slopes and exploratory drilling along the cutoff wall alignment.

• Construction of cutoff wall concrete guidewalls.

• Construction of the 306m long, maximum 23m deep vertical plastic concrete cutoff wall.

•Construction of concrete capping slab over the cutoff wall.

• Grouting of drains, piezometers, sumps and exploratory drillholes.

• Placement of approximately 11,000m3 of RCC with geosynthetic clay liner where required.

• Placement of approximately 55,000m3 of zoned fills to tie the existing till blanket to the new cutoff components, and to provide ballast for the new blanket components.

• Final restoration of the slopes and abutment.

The following sections provide additional discussion on the seepage control plastic concrete cutoff wall and RCC blanket extension.

Construction of plastic concrete cutoff wall

Overview
Although cutoff walls are becoming more common, this project is unique as it is located below a very steep slope within a drinking water reservoir. Construction of the cutoff wall continued 24 hours per day and six-seven days per week over a four-month period from November 2001 to February 2002. Due to concern of potential impact to drinking water quality, the reservoir was off-line and drawn-down during construction. This imposed a strict deadline on construction – be complete in time for spring freshet to refill the reservoir for the peak summer water demand.

A total of 50 panels, ranging from 3m to 8.8m long were excavated using cable clamshell buckets operated from one of two excavation cranes operated by Petrifond Foundations with Vancouver Pile Driving. Typically, the excavation cranes continued excavation at separate panels, with a third service crane used to assist in tremie placement of concrete once excavation was complete.

The cutoff wall penetrated a minimum of 4m into the silt aquitard at the base of the upper aquifer, in order to meet the allowable hydraulic gradient criteria and satisfy construction precedents. Using this minimum penetration depth as a guide, the cutoff wall depth varied from 6m to 23m below the el. 126m construction bench. The total length of the cutoff wall was 306m, comprised of a 46m dog-leg connection with the original till blanket and a 260m length below the blanket extension. The 46m connection included a 36m long right-angle tie-in through the existing till blanket.

Plastic concrete was chosen as the impervious backfill material, based on review of published case histories that indicated a lab permeability of 1×10-9m/s was achievable. In-situ permeability was expected to be about one order of magnitude higher due to scale effects and possible minor construction imperfections. A minimum cutoff wall width of 0.8m was selected to satisfy the conductance design criteria.

Construction sequence and panel details

Once the el. 126m bench was constructed, guidewalls were built using steel reinforced 25 MPa (3600 psi) concrete, to about 1.2m depth and width. Below the inside (near-slope) guidewalls, 100mm diameter perforated drainpipe was placed in gravel drain rock wrapped in non-woven geotextile – connected to 600mm diameter vertical sumps at about 30m intervals to provide a cutoff wall dewatering system. The dewatering system was intended to keep the groundwater level at least 600mm below the top of the guidewalls or slurry level during panel excavation until concrete backfill was complete.

A series of exploratory drill holes were advanced to verify the planned depth of the cutoff wall. These drill holes were grouted after completion as they extended below the final planned depth of the cutoff wall.

The cutoff trench panels were then excavated to depths of between 6m and 23m, using nominal 760mm wide clamshell buckets (nominal bite length of 2.5m). Clamshells with teeth were used for general excavation, switching to smooth-edged cleanup buckets for the final cleanout. The cleanup bucket width was 810mm.

The 50 panels were excavated as ‘opening’, ‘running’ or ‘closing’ panels. Opening panels were the first excavated and were 6m long plus 0.8m on each end for pipe joints. Running panels, extending off an opening panel, were 6.8m long including one new pipe joint allowance, and closing panels were nominally 6m long. Each panel was checked for width, verticality, cleaned joints, depth, length, and slough. A rectifying tool (of 850mm thickness) was passed through each panel, prior to placement of concrete, in order to verify the minimum 800mm width was achieved. Verticality was measured by hanging a plumb bob or joint pipe down the joint locations and recording vertical offsets. Joints were scraped cleaned with the rounded end of the rectifier, and depth was checked by sounding. Slough in the panels was removed by airlift or clean up bucket. Width checks were conducted a maximum of four hours prior to commencement of plastic concrete backfill placement, and a final depth check occurred within 15 minutes of start of backfill.

As excavation proceeded, bentonite slurry was added to the excavation to provide support. Slurry levels were maintained as high as possible to increase stability of trench walls, but always a minimum of 600mm above the local piezometric level, checked against sumps and local piezometers. Slurry was mixed on site in a local tank farm with a slurry delivery and return pipelines located along the top of the guidewalls. Desanding allowed a batch of slurry to be re-used up to three times.

Each panel was logged for general stratigraphy by viewing cuttings in the clamshell, combined with a sounding line for depth. Where necessary, a chisel was used to trim boulders or plastic concrete from adjacent panels at depth. After the target depths were achieved, a single SPT sounding was conducted as a final confirmation of silt below the bottom of panels. The cuttings, contaminated with bentonite and used slurry, were temporarily stored on site then transported off-site (out of the watershed) for disposal.

Prior to start of concrete placement, 760mm diameter steel joint pipes were placed at each end of the open excavation to create a semi-circular joint to tie subsequent running panels to opening panels. Concrete was placed by tremie method with two steel pipe tremie lines per panel. A go-devil chased slurry out of the pipe in advance of the first load of concrete. Depth to top of concrete was measured at each end of the panel and between each hopper just prior to and just after each truck load of concrete was placed. The joint pipes were ‘cracked’ within six hours of completion of the pour, then lifted an additional 0.5 to 1m per hour for the next six hours and were generally removed from the panel 12 to 18 hours following completion of the pour.

Following the completion of the vertical plastic concrete cutoff wall, the top surface was prepared by vacuum removal of laitance and trimming the top portion of set concrete with a mini-excavator. Adeka waterstops were placed along the top of each guidewall, and three more waterstops along the plastic concrete surface. Ribbed PVC waterstops were installed immediately following the placement of the plastic concrete at approximately 21m spacing, in line with the planned contraction joints in the overlying concrete cap and RCC. In final preparation for the RCC, an unreinforced concrete capping slab (typically 30 MPa (4350 psi)), of about 3 to 3.5m width and 1 to 1.5m thickness was constructed.

Stability analysis and monitoring

Klohn Crippen conducted a three-dimensional stability analysis of the open cutoff wall trench to set certain critical design and construction parameters and to confirm its feasibility for construction. The analysis indicated that the wall could be safely constructed at the base of the steep slope by: maintaining a specified density for the bentonite slurry; maintaining the slurry level at least 600mm higher than the adjacent groundwater level outside of the cutoff wall excavation; limiting the width of each panel to a maximum of 6m adjacent to the slope; and maintaining a minimum separation of 30m between excavated or recently poured panels.

As each panel was excavated, the differential between bentonite slurry level and groundwater was maintained by groundwater dewatering or adding additional bentonite, with groundwater levels monitored by piezometers.

Plastic concrete mix design

Mix design requirements were based on required unconfined compressive strengths of 0.75 MPa (109 psi) and 1.5 MPa (218 psi) at seven and 28 days, respectively. A minimum axial strain of 5% at failure in triaxial compression was required to provide ductility during construction and accommodate small strains under seismic loading. The concrete also had to achieve a maximum hydraulic conductivity of 1×10-9m/s measured in the axial direction in a confined triaxial apparatus.

The Engineer reviewed the Contractor’s mix designs prior to start of placement. The initial mix did not meet the required ductility and a series of trial mixes was conducted, varying bentonite/cement and water/cementitious ratios, and assessing variations in strength and ductility using rounded versus crushed coarse aggregate. Mixes were found to have similar behaviour using either rounded or crushed aggregate. The accepted plastic concrete mix consisted of the following yield corrected components per m3:

• 143.6kg cement (Portland Type 10).

• 36.6kg bentonite (un-treated Wyoming).

• 415.9kg water (from Grouse Creek).

• 707.1kg coarse aggregate (CAN/CSA-A23.1-M maximum 20mm).

• 707.1kg fine aggregate (CAN/CSA-A23.1-M).

The bentonite content of the plastic concrete mix design was further adjusted to 24kg/m³ during construction to achieve the design strength. The plastic concrete was batched from a plant set up within the watershed and delivered to the construction bench using four open-box Agitor mixer trucks.

Quality control
The Contractor and the Engineer maintained detailed quality control and quality assurance records respectively.

Records and checklists were maintained for each panel including dimensions, excavation details, stratigraphy, slurry quality, plastic concrete batch slips, and placement summaries. An overall placement graph showing volume of plastic concrete with depth was prepared for each panel. The actual placement graph was compared to the theoretical depth/volume for each panel to check for slough or other potential problems with placement.

Bentonite slurry for trench stability used Bara-kade 90, a polymer treated bentonite product. Fresh slurry density generally ranged from 1042 to 1060kg/m³. Slurry densities in excavations increased to about 1060kg/m³ to as much as 1200kg/m³. The slurry was tested for density, pH, sand content and viscosity a minimum of twice daily, sampled from top, middle and bottom of each open panel. Slurry with a density greater than 1106 kg/m³, sand content greater than 5%, or Marsh funnel viscosity of more than 45 seconds was replaced with fresh slurry prior to approval for concrete placement.

Quality control tests carried out on the plastic concrete during construction indicate the following average properties: unconfined compressive strengths of 0.78 and 1.05 MPa at seven and 28 days, respectively; axial strain in triaxial compression of 10% at failure; and laboratory hydraulic conductivity of 1.7×10^-9 m/s.

In-situ hydraulic conductivity testing was conducted in selected panels, in pre-formed holes, created by inserting steel pipe into the fresh plastic concrete prior to initial set. Initially, coring was attempted but core recovery was poor and highly disturbed due to the low shear strength (0.75 MPa (109 psi)) of the plastic concrete. Coring was discontinued over concerns of potential fracturing of the panel along with the generally poor coring results. Hydraulic packer conductivity testing was conducted in the formed and cored holes and indicated hydraulic conductivity values of 1 to 3 x 10-8m/s, generally an order of magnitude greater than the lab results.

The use of natural gamma downhole logging in cored holes was considered as a means of checking panel widths and for soil inclusions in the plastic concrete due to caving during concrete placement. Calibration trials indicated an insufficient gamma contrast between the plastic concrete and the native soils, and therefore no geophysical testing was conducted.

A detailed review of concrete batch slips was conducted for approximately every tenth panel. Field mixes were checked against the design mix by comparing proportions of mix constituents as a ratio of cement content. The field mixes were found to exceed specified tolerances from the mix design, with, in particular, the amount of bentonite being low, due to slight variations in the density of the bentonite slurry used in the concrete mix. However, since the plastic concrete met strength, ductility and hydraulic conductivity requirements, the variation in field mix to design mix was accepted.

Problem panels
The first (Panel 1) and the last (Panels 47 to 50) of the plastic concrete cutoff wall panels proved to be the most problematic.

Panel 1, a relatively short panel of 6m depth, was the first panel backfilled with plastic concrete. This panel was located at the northern (furthest) extent of the new blanket and cutoff extension at a location where the vertical exposure of the upper aquifer was only about 2m. The specified plastic concrete tremie placement procedures were poorly adhered to and there was some question as to whether the tremie seal had broken during the pour. On this basis, this panel was initially rejected, but subsequent review of cores from the panel and in-situ hydraulic conductivity testing provided adequate basis to accept the panel. Plastic concrete delivery and tremie methods were significantly improved on all subsequent pours.

Panel 50 (and in fact Panels 47 through 50) were located on the 36m long ‘dog-leg’ tie-in through the existing till blanket and were the last to be completed. These panels were located on a temporary berm constructed to allow access to excavate these panels. However, the loosely backfilled temporary berm proved unstable during the initial stages of panel excavation. Bentonite slurry loss into the reservoir occurred requiring emergency spill control response. Slurry loss was controlled by pumping down and/or backfilling the partially excavated trenches. Panels 47 to 50 were eventually completed using a combination of backfilling with lean concrete then excavating through the mix, and limiting the trench lengths to 3m. In the end, the final 3m of Panel 50 was not completed, but an adequate tie-in to the existing till blanket was verified by completing several drill holes, subsequently backfilled.

Construction of RCC blanket

Overview
The use of RCC for a seepage control blanket is believed to be a new application of this type of concrete and composed RCC placed in a minimum 3m horizontal width (considered the minimum constructible width for high volume placement with conventional equipment) from the top of the concrete slab at el. 127m to above the exposed upper aquifer at el. 138m. Additional RCC varying in width from 3m to 1.2m was constructed above el. 138m to tie the RCC into suitable low permeability natural soils.

The RCC blanket was required to achieve the following design criteria: a mass permeability of 5×10^-8m/s in order to meet the conductance design criterion; an in situ permeability of 5×10^-9m/s to account for scale effects and possible minor construction imperfections; and limit the hydraulic gradients in the till aquitard (overlying the Upper Aquifer) to a maximum of four.

Mix design, placement technique and joint detail construction were refined through the construction of three RCC field test strips. A total of 11,000m³ of RCC was placed in the construction of the seepage blanket in 61-lifts of nominally 300mm thickness (post-compaction), over a period of four months, with bedding mix between all lifts. Vertical construction joints, were constructed coincident with the joints in the unreinforced concrete cap located between the vertical cut-off wall and the RCC blanket to control thermal cracking. Contraction joints at the face of the RCC blanket were sealed with a combination of mastic, Volclay tubes, and a geosynthetic clay liner cover.

Coarse and fine aggregate for the RCC was processed from sand and gravel excavated from the east abutment with processing comprising both crushing and screening.

This atypically thin blanket required intense quality control measures and inspection to ensure that no substandard zones were constructed.

Construction sequence and details
Trial RCC mix designs commenced in December 2001 and continued until the start of production RCC in the second week of February 2002. Three test strips were constructed to review and refine the Contractor’s mix design and construction methodology prior to start of production of RCC. The test strips were observed during placement and then cored and sawcut to allow examination of the RCC mass for segregation and quality of both hot and cold joints. The Contractor used the test strip production to optimise required compactive effort and to experiment with contraction joint installation and other construction methods. The test strips and test mix production were also used to calibrate a Vebe (or vibration table) apparatus constructed by the Contractor’s testing agency. The Vebe test was intended to provide field quality control of the RCC consistency and paste. The field tests of the Vebe apparatus proved inconclusive and no definitive calibration was possible, with times varying from 15 seconds to over three minutes. A recurring problem was the bridging of coarse aggregate preventing paste migration. The Vebe test requirements were subsequently removed from the test programme and replaced by a modified Vebe test where a Kango vibratory tamper was used to consolidate RCC in a test cylinder and then the density was measured.

Production RCC was placed between 11 February 2002 and 1 May 2002. RCC was batched at the same batch plant used for the plastic concrete, located at Grouse Pit about 2km from the damsite, and transported by tandem trucks to the East Abutment. RCC was dumped in a transfer box and bucketed to the working surface (cap slab or previous RCC lift) by excavator then spread to its pre-compaction lift height using a D6 dozer. The excavator mixed the RCC from the outer edges to the center of the transfer box to mitigate any segregation caused by dumping from the trucks. Negligible segregation was observed on the placement surface during lift spreading.

Prior to placing the RCC on the working surface, the surface (either concrete cap slab, previous RCC lift or shotcrete surface) was cleaned using high-pressure air and/or high-pressure water and the surface was moistened. Bedding mix was then hand swept or raked to about 10 to 20mm thickness on all lift joints (hot or cold) and along the shotcrete face to ensure paste coverage along potential seepage paths, as the key requirement of the RCC blanket was low permeability. The bedding mix was batched in the same plant as the RCC at Grouse Pit.

The RCC was placed and spread in nominally 400mm thick lifts, then bladed by bulldozer to about 350mm thick lift. Each lift was then roller compacted to 300mm nominal thickness using an Ingersoll Rand double drum roller compactor (IR DD-90) rated at 9.1 metric tons static weight. Plate tampers were used along lift edges. Survey control for the first several lifts comprised level and tape measure, after which the Contractor used laser-levels, both handheld and mounted on the bulldozer blade.

During, and following each day of placement, contraction joints were formed, and the final lift of each day was water-cured and covered with tarps. Exposed lift surfaces were water-cured for about 14 days. Contraction joints in the RCC were aligned with the contraction joints in the concrete cap slab. These joints were designed to initiate cracking from thermal contraction of the RCC at predetermined locations where seepage control measures (mastic and bentonite sheets) were installed.

The joints in the RCC were formed by insertion of hot-dipped galvanised steel plates, 6mm x 400mm x 275mm into each lift of fresh RCC at the specified joint location. Contraction joint plates were inserted by a pneumatic jackhammer with a C-notch attachment, and then the surface of the lift was compacted with a plate tamper or roller compactor. Contraction plates were staggered in a checkerboard pattern on alternating lifts. The spacing between plates created a more tortuous seepage path than if plates had been continuous.

The surface exposure on the outside face of the contraction joint was prepared by trimming fresh RCC overbuild with an excavator clean-up (smooth edge) bucket for a 1m to 1.2m wide strip centred on the contraction plates. The trimmed face was then compacted with plate tampers, and levelled with Sikaflex 2C-NS mastic as necessary. A groove of 100mm depth by 50mm width, aligned with the contraction plates, was sawcut into the compacted face. The base of the groove was tamped flat and a 50mm wide strip of ‘Foampak’ polystyrene bondbreaker was placed along the base of the groove. The bottom half of the groove was infilled with Sikaflex 2C-NS and the upper half filled with a Volclay Hydrobar Tube. The Volclay tubes contain sodium bentonite clay wrapped in water soluble film comprised of polyvinyl alcohol. The infilled groove was then overlain by two bentonite sheets each tacked on one side. Finally compacted granular fills were placed over the bentonite sheets.

During the RCC placement, propagation of cracks was noted only at the design contraction joint locations, except for one crack on RCC Lift #1 which was directly over one of the cracks in the cap slab.

RCC mix design
A cementitious materials (cement and flyash) content of 185kg/m³ was originally selected based on the upper bound 95% confidence limit published by Dunstan (1994) for in situ permeability testing completed on 49 completed RCC structures. The high cementitious content falls in the high paste category of RCC mixes where mix design is normally based on the ‘Concrete Approach’. Flyash was used in the RCC to minimise the amount of Portland cement required and to maximise the paste content of the mix. This improved the coverage of aggregate with paste and lowered the permeability of the RCC.

Final proportioning of the RCC mix was carried out in mix design studies completed by the contractor. The RCC mix design was slightly adjusted three days after construction commenced to increase the paste content and slightly reduce the 28-day strength. The final mix design (RCC-2R2) consisted of the following yield corrected components per m³:

• 93.4kg cement (Type 10).

• 114.2kg flyash (Type CI).

• 116.2kg water (from Grouse Creek).

• 1416.6kg coarse aggregate (excavated and processed from the East Abutment).

• 762.8kg fine aggregate (excavated and processed from the East Abutment).

Mix RCC-2R2 provided a theoretical air-free density (TAFD) of 2503kg/m³, and 97% of this TAFD (i.e., 2428kg/m3) was the specified minimum field density for the RCC.

The aggregate gradations using sand from on-site sources did not fall precisely within the CSA gradation specification, as the sandier fraction, 0.3mm to 5mm slightly exceeded CSA A23.1 limits. This was accepted during the mix design process, as the sandier aggregate increased the mortar content of the mix

The following average RCC properties were achieved during construction: unconfined compressive strengths of 3.2, 12.5 and 15.1 MPa (464, 1,813, and 2,190 psi) at 1, 7, and 28 days, respectively; compacted density of 97.7% of theoretical air free density.

Two bedding mortar mix designs were evaluated, one with coarse aggregate and one with only sand aggregate. Mix BM-2R, using only sand size aggregate, was selected for use. Since only sand aggregate was used, the mortar was spread at about 10 to 15mm thick instead of the original specified 20mm minimum. The bedding mortar mix was nominally 358kg/m³ cementitious comprised of 217kg cement and 141kg (311 lbm) flyash.

Quality control
A detailed Quality Control (QC) and Quality Assurance (QA) testing programme allowed effective checking of design compliance as well as feedback and revision of design components as the work progressed. In particular, monitoring the workability of the RCC allowed revisions to optimise the mix design, as the RCC was sensitive to the aggregate moisture content.

Compressive strength and joint bond strength did not control the stability of the blanket extension as the RCC mass with no bonding of joints was much stronger than the very dense soils. However, compressive strength was considered a useful quality control indicator and a minimum unconfined strength of 15 MPa (2175 psi) at 28 days was selected for construction.

RCC was tested for slump, temperature, unconfined compressive strength, modulus, Poisson’s Ratio, modified Vebe and density, aggregate gradation, and aggregate moisture content at the on site laboratory. Mortar was tested for slump, unconfined compressive strength and aggregate gradation.

Quality Control testing included four UCS tests per shift and hourly density measurements. RCC was also tested at the east abutment by nuclear densometer for in-situ density and moisture, at minimum 30m spacing on each lift. The in-situ density test results indicated the all RCC lifts met compaction specifications.

The bedding mortar field mix (after yield correction) exceeded the original specified cementitious limit of 355kg/m3, with a range of about 380 to 410kg/m³. This specification was changed during the trial pad process, as the extra cementitious content (generally flyash) was required to provide sufficient paste in the mix to fill interstitial voids between the aggregate and produce a paste rich surface promoting good bonding with subsequent lifts.

Quality assurance testing was conducted by Klohn Crippen and by GVWD’s testing consultant. Klohn Crippen conducted in-situ density measurements using a nuclear densometer, and GVWD’s testing consultant conducted periodic sampling of the RCC and aggregate for UCS, slump, temperature, modulus, Poisson’s Ratio, and density (modified Proctor, plus in-situ density by sand cone and nuclear densometer methods).

Selected batch slips from each day of RCC production were reviewed and summarised for both RCC and cement mortar and checked for compliance with specified tolerance of aggregate, water and cementitious constituents with the mix design and were generally within 1% of design for aggregate and cementitious content, and between 1% and 5% for water.

RCC Issues

Thin RCC transition with main RCC blanket
The thin RCC was built with a transition zone from the main RCC blanket to reduce the potential for cracking due to stress concentrations at the contact. The transition from 3m horizontal width to 1.2m horizontal width occurred over approximately 1.5m vertical height, or four lifts. Each lift was stepped in by approximately 350mm to complete the transition

The thin RCC was specified to be a minimum of 0.6m thick measured normal to the slope, which is equivalent to a minimum horizontal width of 1m. The Contractor elected to increase the minimum compacted width to 1.2m with about 300mm overbuild to allow safe passage of roller compactors on the thin RCC.

Two smaller roller compactors were used for the thin RCC: a Bomag Double-Drum 2.5 tonne roller compactor was used for the transition from 3m to 1.2m horizontal width; and a smaller Ingersoll Rand DD-16 double-drum roller compactor, with a roller width of 1m was used for the remaining 1.2m wide sections. As with the full width RCC, the edges and contraction joint faces were compacted using plate tampers. The contraction joints in the main RCC blanket were extended vertically into the thin RCC blanket extension and transition zone.

Hydraulic conductivity tests
Falling head hydraulic conductivity tests were conducted in three RCC coreholes to assess the in-situ hydraulic conductivity of the RCC blanket. Each hole was cored at 130mm diameter to about 1050mm depth. The results ranged from 6.1×10^-9m/s to 6.7×10^-9m/s with a geometric mean of 6.4 x 10^-9m/s.

Differential settlement
During construction, a concern was raised relating to differential displacements developing between the RCC and the natural soils under reservoir loading resulting in a seepage path at the base of the RCC. Although such a scenario was evaluated to have a low probability of occurrence, construction of a low permeability geosynthetic clay liner (GCL) barrier was recommended.

Bentomat ST GCL was placed on the slope 1m below the top of the RCC, where the seepage control blanket did not extend above full reservoir supply level. Bentomat ST is a reinforced GCL comprised of a layer of sodium bentonite between woven and non-woven geotextiles, which are needle punched together.

The GCL was placed horizontally along the slope using a customised backhoe spool roller. The GCL was placed with the woven geotextile side in contact with the soils. Joints at roll ends were overlapped a minimum of 0.6m, and the GCL was nailed to the excavation slope to hold the material in place prior to fill placement. The GCL was covered by plastic or temporary fills to prevent premature hydration by rain. The protective cover was removed and the GCL was then covered with RCC, or, Zone 2 fill where the GCL extended above the RCC. RCC bedding mortar was applied to the Bentomat prior to RCC placement. The Zone 2 fill gradation was designed as a fine filter ‘crack stopper’, in accordance with conventional earthfill dam practice, to plug any cracks that may develop along the RCC/GCL/native soil interface. The Zone 2 fill within 0.5m of the GCL was placed in thin lifts and compacted with hand-operated equipment to avoid damaging the GCL.

Construction issues

Environmental and social issues
The work site is located on the upstream slope of a drinking water reservoir within the Capilano Watershed and therefore construction posed significant environmental and social challenges which could be minimised if the blanket was constructed over one construction season. Although the reservoir was off line for drinking water supply during construction, it did provide water supply for the fish hatchery and Lower Capilano river downstream of the dam. Innovative public impact mitigation measures reduced disturbance to nearby residents, the environment, and most importantly, to the drinking water supply. Measures were taken to reduce off-site traffic, reduce sound and emissions, and maximise re-use of excavated material in the abutment. Environmental monitoring of fish habitat and water releases along with noise and vibration monitoring were conducted to minimise impacts. The construction site was restored using natural materials including replanting over 1000 trees of various native species.

Flooding of the construction bench
Water supply requirements dictated that the reservoir could not be drawn down until the end of the seasonal high demand period in September. Refilling had to commence by the following April in order to achieve a full reservoir at the end of June for the high summer demand period. The construction period below full reservoir level was therefore limited to one fall to spring construction period. The capacity of the low and mid-level outlets at the dam is limited at lower reservoir levels, and the possibility of high precipitation events resulting in reservoir inflow exceeding the discharge capacity was predicted with the potential for uncontrolled reservoir rise, while work was ongoing in the reservoir. A construction bench at el. 126m was selected to provide some flood protection for the cutoff wall construction. Historical precipitation records suggested that flooding of the construction bench could occur on average about once during the construction period.

An evacuation protocol and a detailed precipitation and reservoir inflow protocol were implemented to forecast when the site should be evacuated. Customised weather forecasts were provided by Environment Canada several times a day to prepare inflows and reservoir level forecasts, and, if necessary, flood alerts were issued to the Contractor. The protocol included prescriptions to temporarily backfill all open trenches containing bentonite slurry materials, and remove all working equipment and machinery from the Construction Bench. Three flood alerts required evacuation – twice flood waters came within 0.5m of the construction bench, with flood waters cresting at 4.3m above the construction bench on the third event. In these three cases, the evacuation protocol minimised the risk to personnel, equipment, and the potential environmental impact of spills in the reservoir. There were no adverse environmental impacts during any of the flood evacuation events.

Partially excavated panels were backfilled with till, sand or gravel as part of the evacuation protocol to maintain slope stability under flood and subsequent drawdown conditions. As flood waters receded, backfilled working panels were re-excavated and a detailed surveillance programme, including piezometric levels, seepage, survey stakes and monitoring pins was conducted as a check for any signs of instability. There were no post-flood stability problems.

Slope stability and restoration
The benches on the upper 50m of the excavated slope above full reservoir level were graded to provide drainage. Landscape design included vegetative cover with inclusion of live pole drains, fascines, wattles, coarse woody debris and hydroseeding. The design of the lower 20m of the excavation slope, below full reservoir level, included fibre-reinforced shotcrete to minimise slope ravelling and erosion while work was being conducted on the construction bench.

Conclusions

The postulated failure mode for the east abutment was development of an erosion pathway from the end of the original blanket to the T505/506 drainage tunnels. The blanket extension construction has met the design requirements. In particular, the abutment materials considered most susceptible to internal erosion were covered by non-erodible RCC and plastic concrete materials. The overall seepage path length has also been substantially increased by the blanket extension, which reduces the risk of forming a new erosion pathway.

Groundwater modelling provided a method to extrapolate measured piezometric levels to areas important for design but difficult to measure. One such important area was near the end of the original blanket. Recalibration of the groundwater model to post-construction conditions indicates that the blanket extension modelled conductance is similar to the modelled conductance of the original blanket, which has performed well for 50 years.

This well-managed project, which was completed on schedule and within budget, has significantly improved public safety and the long-term stability of the dam.

Author Info:

Neil Singh, P.Eng, Ryan Douglas, P.Eng., PE, and Steve Ahlfield, P.Eng are with Klohn Crippen Consultants, Vancouver, BC, Canada; Murray Gant, P.Eng is with Greater Vancouver Regional District, Burnaby, BC, Canada.

This paper was originally presented at Dam Safety 2005, the annual conference of the Association of State Dam Safety Officials (ASDSO). Thanks are given to ASDSO for kind permission to republish the paper. For more details visit www.damsafety.org