The Seymour Falls dam is a key element in Canada’s Greater Vancouver Water District’s (GVWD) network of three watersheds. The 45-year-old composite earthfill embankment and concrete slab and buttress dam was found to not meet current seismic design standards, so upgrades to meet the requirements of the Maximum Credible Earthquake (MCE) are necessary. Part one of our two part report describes the structure and looks at how the safety problem was identified and analysed through site investigation and liquefaction assessment
The Greater Vancouver Water District (GVWD) provides a source of drinking water to over 2M people in its 18member municipalities. This includes acquiring and maintaining the supply, treating it to ensure its quality and delivering it to the municipalities. Water is collected from three mountainous watersheds: Capilano, Seymour and Coquitlam and is delivered by an extensive system of 22 reservoirs, 15 pumping stations and over 500km of supply mains.
The Seymour Falls dam provides approximately one-third of the total regional water supply and is a key element in the system. As shown in Figure 1, the dam is located on the Seymour river, approximately 18km north of the Burrard inlet and within the Lower Seymour Conservation Reserve (LSCR). Shown as the hatched area in Figure 1, the LSCR is open to the public for recreational activities, including cycling, hiking and in-line skating, and education programmes. A salmon hatchery is located approximately 300m downstream of the dam. This environmentally and socially sensitive setting poses significant challenges that are being effectively met during the ongoing seismic upgrade construction for the dam.
In the late 1920s, a 6m high concrete dam was constructed at the existing Seymour Falls dam site to supply water to the Vancouver area. With the rapid population growth in the Lower Mainland, the GVWD decided in the 1950s to build a new higher dam to provide additional water storage. The dam was originally designed to have an ultimate height of about 47m above the riverbed. In the early 1960s, Stage 1 of the dam was constructed to a height of about 30m with provision for future raising – Stage 2, which has not been built. A photograph and the layout of the existing dam, prior to the current construction phase, is shown in Figures 2 and 3(a). Figure 3(b) provides a schematic summary of the proposed seismic upgrade. The dam is a composite structure consisting of a 235m long slab and buttress concrete section, a concrete gravity retaining wall, a 220m long earthfill embankment and an extensive upstream impervious earthfill blanket.
The Seymour river runs in a valley that has experienced many periods of glaciation. Following the last glacial retreat, about 10,000 years ago, land levels rose relative to the sea level. Sediments were rapidly deposited in the Seymour valley by erosion of material from the valley sides and alluvial material transported down the valley. As the land rose and the thickness of valley bottom deposits increased, rapid deposition of material from the valley sides dammed the river valley in several locations creating lakes. Periodically, such lakes were formed at the existing dam site by a debris fan known as the Cougar Creek fan. In both the marine and fresh water depositional phases, the Cougar Creek fan was deposited contemporaneously with fine-grained, deep-water marine and shallower water lacustrine deposits, hence the northern distal lobe of the fan is inter-fingered with fine grained lacustrine sediments.
The Cougar Creek fan, which now underlies the west side of the dam site, covers a semicircular area of radius about 800m. The earthfill dam is built from the west valley wall out overtop of the Cougar Creek fan, an area that includes a buried valley or thalweg. On the opposite (east) valley wall is a narrow bedrock spur which projects to the edge of the Cougar Creek fan. The upper surface of the bedrock spur is approximately level with the riverbed, and the concrete portion of the dam is founded on this bedrock outcrop. This bedrock spur falls gradually to the west below the Cougar Creek fan to about 60m below ground surface, before being truncated by the old rock valley thalweg, which is infilled by over 150m of Cougar Creek fan deposits. The thalweg is immediately below the right abutment of the earthfill dam. Bedrock then climbs steeply to form the west valley wall above the Cougar Creek fan.
The majority of the bedrock foundation for the concrete structure is composed of a hard, fresh to slightly weathered granodiorite. Locally, andesite dikes bisect the rock mass, possibly as a result of an intrusion following an ancient shearing event. This shearing event resulted in a reduction of quality of the surrounding andesite and granitic rock, characterised by closely spaced, slightly open, fractures. The primary discontinuities within the bedrock foundation consist of two orthogonal joints with a lesser-developed subhorizontal joint, which may have developed in response to post-glacial rebound. Also, there are a number of secondary joint sets, which is typical for most granitic formations. The bedrock surface is extremely irregular, and is known to include several deep scour holes or pockets of several metres depth and width.
The upper 20m to 40m of the Cougar Creek fan is designated as the upper Cougar Creek fan deposit and contains some very loose granular material. The surficial 18m of the upper Cougar Creek fan deposit is extremely coarse bouldery material, comprised of boulders, cobbles, sand and gravel. In parts of the fan where the bedrock is relatively deep, the material becomes progressively finer below the bouldery layer, transitioning to coarse sand at about 30m depth. More dense preglacial lower Cougar Creek fan deposits are found below. Upstream of the dam, a distance of about 100m into the reservoir, the upper Cougar Creek fan deposits transition to predominantly lacustrine silt.
Groundwater levels at the site are heavily influenced by both the reservoir and local fan recharge. The reservoir is contained by a partially effective impervious blanket – changes in reservoir levels are directly measurable by changes in under-seepage and piezometric levels in the fan. In addition, precipitation events and snow melt directly recharges the local catchment, primarily the Cougar Creek catchment area, but also adjacent hydro-geologically connected areas. Generally, groundwater flows south and east below the existing dam around the buried bedrock spur and then to the Seymour river.
Description of the existing dam
The earthfill section consists of three major elements: the main embankment; the land blanket; and the lake blanket shown schematically in Figure 3(b). The main embankment consists of six major zones – a typical cross section is shown in Figure 4. The impervious section of the embankment is an inclined central core composed of compacted clayey silt. Upstream of the core, a transition section of sand and gravel was constructed between the core and the granular shell. The upstream shell is a well-graded mixture of sand, gravel, cobbles and boulders; the core extends below the upstream shell and is connected to an impervious blanket of clayey silt at the toe of the dam. The impervious blanket extends about 200m into the reservoir to tie into rock and into the natural lacustrine silt, and the impervious blanket is 1.5m thick and was constructed in two phases consisting of a ‘lake blanket’ in the pre-1960 reservoir and a ‘land blanket’ over the Cougar Creek fan sediments. The blanket is not a perfect cut off and there are windows at the upstream end, which allow significant seepage to pass beneath the dam. Downstream of the core are fine and coarse filter zones of clean, well-graded sand, and clean gravel, respectively. The downstream shell consists of pit run gravel and sand with some cobbles.
The concrete section consists of a slab and buttress dam, comprised of a series of upstream sloping concrete slabs supported on buttress walls spaced at 6.7m centres. Discharge facilities include twelve 6.1m wide overflow spillway bays with a total width of 73.2m, two 1524mm diameter outlets fitted with Howell Bunger valves and one 610mm diameter low level outlet. At the west end of the concrete section, the transition block connects the concrete section to the earthfill embankment. Downstream of the transition block, the embankment fill is retained by a mass concrete gravity wall (GRW). The dam also includes an intake tower and a small hydroelectric turbine, which supplies power to the water treatment facility and the Seymour Fish Hatchery. The entire concrete section, including the transition block and the concrete gravity wall, is founded on bedrock.
A 2390mm diameter steel pipeline (Main #2) connects the reservoir to the GVWD water distribution network. Currently, water is treated at a chlorination plant located at the toe of the gravity wall, but the GVWD is in the process of constructing a major water filtration facility approximately 12km downstream of the dam, which will treat water from both the Seymour and the Capilano reservoirs. The chlorination building and Main #2 are located just south and east of the gravity wall (Figure 3).
Dam safety assessment
Historically, it is known that moderate earthquakes periodically occur in the coastal regions of southwestern British Columbia including the Lower Mainland. However, during the 1980s, studies indicated that the potential for a moderate or large earthquake in the region is much greater than previously anticipated. Even though seismic events are over relatively quickly, the loss of lifeline infrastructure such as water supply dams can cause not only major public health and safety problems, and disrupt normal activity in the region for weeks or months, but also result in many deaths and subsequent property and environmental losses.
Although the existing dam is in good condition, it does not meet current seismic standards. The Provincial Dam Safety Regulation requires all dam owners to be responsible for the safety of their structures and carry out necessary updates to meet current design standards. Due to significant urban development downstream, both in terms of resident population and property development, the Provincial dam safety branch has rated the dam a very high consequence structure in the event of a failure. Based on the Canadian Dam Association’s Dam Safety Guidelines, this will require the dam to resist a Maximum Credible Earthquake (MCE) established by the deterministic method or an earthquake with an annual exceedance probability of 0.0001 derived by the probabilistic method. This is also consistent with GVWD’s internal seismic design criteria, which requires all dams and other important facilities to be designed for the MCE event. The GVWD took the initial steps to upgrade the Seymour Falls dam in the late 1980s by retaining Canadian company klohn-crippen Consultants to carry out engineering assessment and conceptual design. The assessment indicated that the concrete section was more deficient than the earthfill section. A partial upgrade was conducted in 1994 on the concrete section to resist the Design Basis Earthquake (DBE), corresponding to a National Building Code of Canada level of earthquake with a return period of 475 years. At the time, no upgrade was carried out on the earthfill section, as site investigations and analyses indicated that the earthfill was stable under the DBE.
In 1998, the GVWD tendered the detailed design of the MCE seismic upgrade of the Seymour Falls dam. The contracts for the earthfill section and the concrete section were awarded to Klohn Crippen and Acres International Limited, respectively, both of Vancouver, British Columbia.
Concurrent with the detailed design work, the GVWD started negotiating with the Federal, Provincial and Municipal agencies as well as the Seymour Salmon Hatchery to obtain approval for the seismic upgrade. The GVWD conducted a major upgrade to the hatchery infrastructure in 2003 to mitigate potential construction impacts to the hatchery.
Construction poses significant environmental and social challenges at the Seymour Falls dam. The GVWD began extensive public consultation and communication with stakeholder groups in 1998, which has continued through the construction phase. Several measures mitigated social/environmental impact during construction include maximising the re-use of excavated foundation materials for the embankment construction, using gravel pits within the LSCR and batching concrete on site to minimise import of material and hence truck traffic, and completing a new C$3M (US$2.8M) recreational pathway to minimise impacts on the recreational users within the LSCR by moving them off the mainline access road to the dam now dedicated to operations and construction traffic.
The detailed design of the MCE upgrade was completed in 2003. The seismic upgrade construction was tendered in late 2003 and awarded to Peter Kiewit Sons in January 2004. Construction commenced in March 2004 and is expected to be complete in late 2006 or early 2007. The proposed seismic upgrade for the earthfill section is shown schematically in Figure 5.
Seismic hazard assessment
BC Hydro International, assisted by the Geological Survey of Canada (BCHI 1998), evaluated the site seismicity in 1998. The controlling events for the site are local random earthquakes, local deterministic sources, the closest being the Britannia Fault, and deep seated intraplate events beneath Georgia Strait.
Recommended design response spectra for these events for use in dam remediation were:
• M7.5 intraplate earthquake: deterministic 84th percentile.
• M6.5 Britannia Fault local earthquake: deterministic 50th percentile.
• Random local earthquake: Uniform Hazard Response Spectrum (UHRS) with a 10-4 annual exceedance frequency.
Target parameters used for earthquake time history scaling are shown in Table 1.
The time histories were selected based on earthquake magnitude, distance and peak ground acceleration, and then scaled to fit the target spectrum over the full period range as well as to the target values in Table 1. At the onset of the analysis work, it was determined that three earthquake time histories would be selected for each design earthquake source. For those sources where a sufficient number of time histories were not available, the time histories with the best possible fit were selected with interpretations by Idriss (1998), to assist the final selection of records for use in analysis.
Earthfill design criteria
Basic criteria for the upgrade appropriate for the very high consequence structure were selected for safety under the MCE and the Probable Maximum Flood (PMF). Additional project criteria included the requirement for uninterrupted operation of the dam and maintenance of current levels of earthquake and flood protection during construction. Consequently long term safety factors, appropriate to very high consequence dams, were to be maintained at all times. Table 2 shows the key criteria that was were identified.
It was also necessary to monitor and prevent environmental impacts on the salmon hatchery 300m downstream of the dam, which partly relied on water supply from groundwater within the Cougar Creek fan as well as water released from the reservoir.
Foundation and liquefaction assessment
In Situ Penetration Testing and Soil Parameters
For the earthfill dam, the main issue is the liquefaction susceptibility of the coarse granular deposits in the Cougar Creek fan. Extensive site investigations were completed between 1990 and 1998 to assess if the coarse granular Cougar Creek fan deposits were susceptible to liquefaction under design earthquakes. A considerable problem in assessing the liquefaction potential was the difficulty of obtaining meaningful standard penetration test (SPT) values because of the influence of very coarse particles up to boulder size. A typical gradation envelope of the native fan material is shown in Figure 6(a), and a typical boulder is shown in Figure 6(b).
Investigations used conventional SPT, Becker Penetration Test (BPT) and shear wave velocity testing among other methods and considerable effort was made to correlate BPT with conventional SPT, done in accordance with ASTM 1586-84. Both Harder and Seed (1986) and Sy (1997) were used for the BPT/SPT correlations. Eventually it was concluded that, although the Harder BPT to SPT conversion gave results closer to ASTM SPT results than Sy’s method, neither BPT conversion gave a reliable correlation to SPT. Although, no conclusive reason for the poor BPT to SPT conversion was identified, Klohn Crippen believes it was due to formation of plugs of coarse material being pushed in front of the Becker closed end casing. These plugs are not accounted for in the analysis but would form in a highly variable way in the loose, heterogeneous Cougar Creek fan deposits.
After further assessment, it was determined that SPT values obtained in zones of finer material provided meaningful data both for design and post construction improvement measurements. A method of interpreting SPT values based on blow count as 4 x the lowest 3 consecutive blows recorded over 25mm increments, in the penetration interval 0.15 to 0.45m, was developed and found to provide reasonable data in the cobbly ground.
The site investigations provided soil samples in both the existing dam and the foundation soils. Lab testing, combined with emperical correlations of shear wave and SPT values, were used to estimate unit weights and shear strength parameters. Residual strengths were estimated using a stress normalized ratio (Su/_’vo) of 0.06.
Soil liquefaction was assessed using two approaches, firstly using a conventional Seed simplified analysis, as modified by NCEER and reported recently by Youd et al 2000. For this approach SHAKE was used to calculate the cyclic stress ratio and estimate the liquefaction potential of foundation soils. This analysis also provided a required cyclic shear strength and therefore an initial target SPT (N1)60-CS value for ground improvement. Secondly, a self-triggering approach was applied (Beaty and Byrne, 1998) using the 2D finite difference program FLAC. Both total and effective stress methods were analysed in FLAC. The FLAC analysis provided estimates of the earthquake induced deformation of the dam, including the effects of liquefied zones in the foundations and were used to confirm the SHAKE results and as an aid to optimizing the layout and zoning of the remediated dam.
The SHAKE based simplified approach indicated extensive zones of liquefaction with almost all the Cougar Creek fan granular deposits failing to meet the minimum safety factor of 1.1 against liquefaction. The FLAC analysis included an incremental approach to softening of soil by keeping count of the number of cycles to cause liquefaction. Thus soil behaviour in the FLAC model was influenced by material that liquefied or softened earlier in the earthquake. Although FLAC predicted less extensive liquefaction than the simplified method, the consequences of liquefaction still included dam crest settlements of nearly 4m, as well as likely failure of blanket and core possibly leading to piping. A section of the dam showing the FLAC predicted zones of liquefaction are shown in Figure 7.
The FLAC analysis aided in preparing limit equilibrium stability analyses by applying residual strengths in zones where liquefaction was predicted to check post MCE stability and long-term stability factors of safety. As a general observation, the (N1)60-CS required to achieve a safety factor > 1.1 against liquefaction is about 20-blows/0.3m, which was about twice the pre-improvement site average blow count.
The authors are Len Murray, P.E. P.Eng and Neil K. Singh, P.Eng of Klohn Crippen Berger Ltd., Vancouver, BC, Canada; and Frank Huber P.Eng and David Siu, P.Eng of Greater Vancouver Water District, Vancouver, BC, Canada.
In Part two of this paper, we will be looking at which earthfill section remedial measures took place and examine how ground improvements were implemented.