The findings from a recent seismic stability evaluation of Anderson dam in the US have prompted the initiation of a US$110M improvement project. Frank Maitski and Marc Ryan give more details.
Anderson Dam, constructed in 1950, is located in Santa Clara County, California in the US. It is owned by the Santa Clara Valley Water District and creates Anderson Reservoir, which has a maximum capacity of 111.5Mm3. The reservoir is the district’s largest, providing a significant surface water supply. All district dams are under the jurisdiction of the California Department of Water Resources, Division of Safety of Dams (DSOD), and a small powerhouse (constructed in 1986) puts the dam under the jurisdiction of the Federal Energy Regulatory Commission (FERC). Because of the communities downstream of the dam, the DSOD and FERC have classified Anderson dam as a high hazard structure.
Anderson dam is a zoned rockfill and earthfill dam with a maximum height of about 73m. The crest is approximately 426.7m long and 7.6-13.1m wide. The zoned dam includes upstream and downstream rockfill shells (Zones 1 and 4), a compacted clay core (Zones 2 and 3), and graded transition zones between the rockfill and clay core.
This seismic stability evaluation of Anderson dam was performed to address requests from FERC and DSOD, and to satisfy the requirements of the district’s dam safety programme. The evaluation includes seismic analysis of the embankment during the maximum credible earthquake (MCE) and the effects of potential fault rupture from fault traces in the foundation of the dam. The seismic stability evaluation was performed by AMEC Geomatrix of Oakland, California.
Anderson dam is located on the eastern margin of Southern Santa Clara Valley, at the front of the southwestern flank of the Diablo Range (Figure 1), approximately 2km from the Calaveras Fault and 10km from the San Andreas Fault. The Coyote Creek and Range Front faults are located along the eastern margin of Southern Santa Clara Valley. Anderson dam is situated on the Coyote Creek fault, where the Range Front fault to the south splits from the Coyote Creek fault. These faults likely are part of a single fault system (the Coyote Creek – Range Front [CCRF] fault zone) that merges at depth with the Calaveras fault to the east.
The CCRF faults are considered to be conditionally active per DSOD criteria. Detailed geologic mapping of the site and vicinity was performed as part of the fault rupture hazard evaluation.
The project approach included a review of previous work at the site, performing additional field exploration and lab testing, updating the characterisation of seismic sources and earthquake grounds motions, and then performing the stability analyses.
Field and laboratory data for this study was derived from three main sources: a seismic stability evaluation performed by WA Wahler Associates in 1977, investigations performed as part of the District’s Phase 1 Dam Instrumentation Program in 2007 (Fugro, 2007), and the investigations performed as part of this study. Historical documents, including photographs from the original construction, were essential in putting together the complete construction history of the dam.
The study was performed to evaluate the long term, rapid drawdown, and seismic stability of the dam. The study used two dimensional finite-element analyses to evaluate the static stresses and dynamic response of the dam embankment, and the results of in situ tests to evaluate the cyclic resistance of the embankment and foundation materials.
A representative cross-section of the dam was developed based on the construction history and information collected from field exploration and laboratory testing programmes. The analysis was performed for the normal maximum reservoir level. The phreatic surface within the dam was based on dam piezometer readings.
Horizontal peak ground accelerations (PGA) and response spectral accelerations at the dam site were estimated using a deterministic approach using the MCE on the controlling seismic sources, which include the nearby active and conditionally active faults (ie the San Andreas fault, the Calaveras fault and the Coyote Creek fault). The response spectra for ground motions postulated for the site were estimated at the 84th percentile level since all of the fault sources have moderate to very high slip rates (1 to >9mm/yr) and the consequence of failure of the dam is extreme. The parameters used to develop the spectra are shown in Table 1. The Coyote Creek fault, moment magnitude (Mw) 6.6, controls the spectral accelerations between the PGA and a period of about 1.5 seconds. The spectral accelerations are controlled by the Calaveras fault, having a MCE of magnitude Mw 7 ¼ at a distance of 2km from the site, above a period of 1.5 seconds. Three sets of recorded time histories (each set having two horizontal components) were selected and spectrally matched to represent the ground motions for each of the two scenario earthquakes, resulting in a total of six sets of time histories.
Analysis and results
Two other types of material were identified and analysed in the study in addition to the zoned embankment materials; alluvium in the foundation, and the lowest layer of downstream rockfill material, designated lower finer fill (LFF) in the study. The initial material placed as part of the downstream shell was discovered to be of lesser quality than the remaining rock material above it. This first material from the borrow site was noted to have a finer gradation than the remainder of the Zone 4 material, hence the LFF designation.
The in situ density of the embankment and foundation material was estimated by measuring penetration resistance using the Standard Penetration Test (SPT) or Becker Penetration Test (BPT). SPT testing was performed at 23 boring locations, and BPT testing was performed at 19 boring locations. The SPT and BPT results were corrected using standard methods and evaluated to determine the appropriateness of their use in the analyses. A reasonable correlation between SPT and BPT (N1)60 values was not found, even when considering gravel corrections in the SPT data. Because of this and the high gravel content in the lower finer fill and alluvium, the SPT data was not used to assess the cyclic strength of these materials. The evaluation of cyclic strength was based solely on the results of the BPT test. Figures 2 and 3 show the BPT results for the LFF and alluvium, respectively.
The Zones 1 and 4 rockfill shells of the dam consist primarily of large boulders and cobbles with some sand and clay, and are free draining under all loading conditions. The Zone 2 and 3 core materials are compacted clayey sand with gravel, and are well compacted and not susceptible to liquefaction. The drained and undrained strength properties for these materials were developed from published relationships and the laboratory testing performed on relatively undisturbed samples.
Long term and rapid drawdown stability evaluations were performed using limit-equilibrium slope stability analyses methods through a representative cross section (Figure 4) The long term analysis shows the dam is stable under the current static loading conditions. The rapid drawdown analysis shows that the dam is stable even when the reservoir is lowered rapidly. The Zone 1 rockfill shell on the upstream slope is free draining and provides adequate buttressing of the clayey core under these conditions.
The results of the response analysis were used to estimate the liquefaction potential of the LFF and the alluvium. The LFF and alluvium were found to be liquefiable for all the cases analysed (Figure 5). Post liquefaction stability analyses utilized residual strengths within these liquefied zones. Post-earthquake stability analyses indicate that the downstream slope of the embankment will become unstable during and after earthquake shaking, and that the upstream slope of the embankment may undergo 3.3-7.6m of permanent deformation during and after earthquake shaking, potentially leading to an uncontrolled released of reservoir water. The results of the analyses are consistent with observed past performance of dams with similar foundation conditions subjected to strong ground shaking.
The results show that the presence of the saturated LFF in the downstream shell and the alluvium under the upstream shell, while limited in thickness, have a significant adverse effect on the seismic performance of the dam. The estimated permanent deformation, both upstream and downstream, are considered unacceptable for dam safety.
Water level restriction
Based on the results that the LFF and alluvium were liquefiable, the district authorised additional study to evaluate the relationship between reservoir level and seismic dam performance (deformation) in order to recommend a maximum interim reservoir level that provides an adequate level of safety to the downstream public. The performance of the embankment was evaluated using two-dimensional, nonlinear finite-difference analyses with program FLAC. The FLAC analysis evaluated the performance of the entire dam simultaneously and estimated the deformed shape of the embankment after shaking. The results estimate crest settlements on the order of 4.5-7.3m when analyses assumed an interface layer that liquefies at the start of earthquake shaking and when neglecting model dilation. Allowing for the potential for the development of cracking to depths of about 6.1m based on reported case histories, a water level restriction between 12.2m and 13.7m below the dam crest was recommended as a prudent measure against the potential for uncontrolled release of reservoir water. A restriction of 13.7m below the dam crest was recommended by the District and approved by both FERC and DSOD. The estimated annual loss in water supply yield is 13Mm3 which represents 28% of the annual production of the system.
To correct the deficiencies of Anderson dam, the district initiated a US$110M capital improvement project; the Anderson Dam Seismic Retrofit Project. The objective is to make improvements necessary to stabilise the dam embankment for the MCE, possibly modify or replace the outlet works, and possibly modify the spillway to increase freeboard during a PMF event. Initial planning of the project has begun and the start of construction is scheduled for 2016 with completion in 2018, after the project is fully planned, permitted and designed.
Frank Maitski, Deputy Operating Officer, Water Utility Technical Support Division, Santa Clara Valley Water District, 5750 Almaden Expwy, San Jose, CA, US.
Marc Ryan, Project Manager for AMEC/Geomatrix
Information for this article was extracted from AMEC’s seismic stability evaluation reports.