This article was originally presented at ANCOLD (2017) Conference, Hobart
Haldon Dam is a 15 m high zoned earth-fill embankment irrigation dam, located approximately 10 km south-west of Seddon, in the Awatere Valley, New Zealand. The crest and upstream shoulder of the embankment suffered serious damage during the 2013 Cook Strait earthquakes, and the Regulator enforced emergency lowering of the reservoir by 5.5 m to reduce the risk of flooding to Seddon Township from a potential dam failure.
AECOM was engaged by the owner to carry out a forensic analysis of the damaged dam and subsequently the design of the 2-Stage remedial works. The remedial works addressed the existing dam deficiencies and earthquake damage in order to restore the dam to full operational capacity and gain code compliance certification.
Key features of the approach included holding a design workshop with the owner prior to undertaking detailed design, careful rationalisation of the upstream shoulder to optimise the competing interests of strength and permeability, contractor and regulator involvement in the design and construction process, and balancing risk and constructability with the chimney filter retrofit.
This paper presents a description of, and approach to, remedial works solution undertaken to remediate a substandard and earthquake-damaged dam to fully operational status in an area of high seismicity. Applying this approach, the objective of achieving a robust, safe, economical design that was acceptable to the regulators and the owner was achieved.
Haldon Dam is a 15 m high, Medium Potential Impact Category (PIC) (potential for loss of life, moderate socio-economic, financial or environmental damage) irrigation dam located on Starborough Creek approximately 10 km south-west of Seddon, in the Awatere Valley, New Zealand. A site location map is presented in Figure 1.
Figure 1: Site Location Map
Figure 2: Typical Section of Original Dam Design
Between 2003 and 2008, investigations, design and construction of the original dam was completed. The original design comprised an approximately 15m high by 170m long crest embankment with conditioned mudstone core, alluvial gravel shoulders, and downstream chimney filter/drainage zone, with a storage volume of approximately 250,000 m3, and design flood event of 1 in 1000 Annual Exceedance Probability (AEP). It is founded on in-situ mudstone. The dam has a separate spillway on the left abutment and an environmental flow release pipeline on the right abutment, with an intake at the head of the reservoir. The spillway consisted of an 18.5m long, 5m wide broad crested weir with an armoured energy dissipation zone, and 120 m long riprap lined channel with a second energy dissipation basin at the outlet. A typical section of the original design is shown in Figure 2.
Initial construction was completed in February 2008. During construction there were a number of non-compliant issues with the design and construction which were not closed out at the time. These, and a number of other issues in combination, resulted in poor performance of the dam on first filling, including a longitudinal crack on the downstream side of the crest, and minor slumping of the upstream shoulder. The construction contractor subsequently went into liquidation with the result that no as-built construction records were available.
The dam filled during a single rainstorm event in July 2008 and came close to overtopping when it was found that the spillway had been constructed too high and narrower than the design. Following this, remedial work on the spillway was carried out but the dam was not put into service due to ongoing concerns over its performance.
After damage to the dam by the Cook Strait Earthquakes, the Regulator (Marlborough District Council (MDC)) and its dam consultant (MWH Limited) enforced emergency lowering of the reservoir by 5.5 m to reduce the risk of flooding to Seddon Township from a potential dam failure. This involved excavation through the spillway and spillway channel, and excavation of a second lower level outlet trench through the left abutment.
Following these events, AECOM were engaged to perform an independent forensic analysis of the dam. This involved investigating and defining the deficiencies and advising on remedial solutions, and whether repair was practicable, given the extent of the damage. A remediation solution was identified, and the estimated cost of this was compared to the estimated cost of a full dam demolition and reconstruction. This analysis concluded that remedial works were an economically viable approach.
This paper presents an overview of the remedial works undertaken to restore the dam to full operational capacity, and the rationalisation process followed for each design feature. Specific challenges associated with the dam site such as regional seismicity and sensitive foundation geology will also be presented, and methods employed to overcome these challenges.
Figure 3: Locations and Focal Mechanisms for Cook Strait Earthquakes of 2013
2. Geological Setting
The QMAP Geology of the Kaikoura Area (Rattenbury et al., 2006) indicates that the area around the dam site is underlain by Late Pleistocene-age alluvial deposits consisting of ‘weathered, poorly sorted to moderately sorted gravel underlying loess-covered; commonly eroded aggradational surfaces’. This is underlain by the Miocene-Pliocene age Starborough Formation, consisting of ‘poorly bedded sandstone and sandy siltstone in the Awatere Valley; and siltstone near White Bluffs’. Also on site is the Late Miocene age Upton Formation Siltstone, consisting of ‘Poorly sorted and poorly bedded channelised greywacke conglomerate with lenses of sandstone and sandy siltstone’.
The site is located in the Australian and Pacific plate boundary transition region with the Hikurangi subduction zone to the north-east and continental collision to the southwest. Movement along the plate boundary is dominantly accommodated by the oblique strike-slip, north-east – south-west trending faults in the region. The regional faults that provide the greatest contribution to the seismic hazard at the site include the Alpine, Hope, Clarence, Awatere, and Wairau Faults; all of which generate earthquakes with magnitudes greater than Moment Magnitude (Mw) 7.0, with recurrence intervals of 300 to 2,000 years. No faults are known to intersect the site.
Figure 4: Vertical Settlement of Upstream Shoulder Following the August 2013 Event
Figure 5: Longitudinal Crack Following the August 2013 Event
3. Recent Seismicity
3.1 2013 Cook Strait Earthquakes
The Cook Strait Earthquake sequence of 2013 consisted of two major earthquakes. The main Mw6.5 shock of 21 July 2013, and the secondary Mw 6.6 shock of 16th August 2013. Seismographs in the region indicated that July 2013 event was likely to have induced peak ground acceleration (PGA) at the dam site of approximately 0.21g, while the August 2013 event is thought have induced slightly higher PGA at the dam site. Several significant fore- and after-shocks occurred in association with the main events, ranging from Mw 4.7 to Mw 5.9. Figure 3 shows the locations and focal mechanisms for the Cook Strait earthquakes of 2013.
Following the main earthquake of 21 July 2013, the dam experienced cracking along the crest. These longitudinal cracks were most evident on the upstream side of the crest, with smaller cracks down the centre of the crest. The longitudinal cracks varied in width up to approximately 100 mm with the widest cracking located where the embankment is at its highest.
Following the 16 August 2013 event, the longitudinal cracks had opened further and the crack towards the middle of the dam where the embankment is at its highest had increased up to 500 mm wide. The upstream shoulder appeared to have separated laterally from the core and slumped vertically. This cracking was in the same location as the cracking observed following first filling, indicating that the post-earthquake cracking could have been a continuation of deformation of the upstream shoulder which had initiated prior to the earthquakes. Figures 4 and 5 show the damage to the crest and upstream face of the embankment following the August 2013 event.
Monitoring on 2 October 2012 indicated that post-construction embankment settlement had almost stopped. However, the monitoring results of 19 August 2013 indicated that the embankment settled further (up to 130 mm) due to the main and aftershock sequences. This settlement was concentrated at settlement monitoring point Pin 2 which is located approximately at the highest section of the embankment on the downstream edge of the crest road.
The damage experienced by the dam as a result of the Mw6.5 main shock, is categorised as “major” and the damage following the Mw6.6 aftershock is categorised as “severe” using the system of Pells and Fell (2003).
Following the August 2013 event, the Regulator (Marlborough District Council (MDC)) and its dam consultant (MWH Limited) enforced emergency lowering of the reservoir by 5.5m to reduce the risk of flooding to Seddon Township from a potential dam failure. Seddon is some 12 km downstream, where a peak flow of 120 m3/s were expected (higher but comparable to the 100 AEP flood discharge of 90 m3/s). The lowering involved demolition of the spillway weir and stilling basin and excavation through the spillway channel, and excavation of a second lower level outlet trench through the left abutment.
3.2 2016 Kaikoura Earthquake
An Mw7.8 earthquake occurred on 14 November 2016, initiating approximately 15 km north-east of Culverden, and rupturing in a north-east direction on a number of different faults over a distance of approximately 200 km. The largest energy release was near the township of Seddon, approximately 15 km north-east of the dam site. Peak ground accelerations associated with the Kaikoura earthquake were recorded at 0.75 g (at the Seddon Fire Station) and 1.27 g (at the Ward Fire Station). A significant amount of aftershocks occurred following this event, as shown in Figure 6.
Figure 6: Kaikoura Earthquake Sequence and Aftershock Locations
This earthquake occurred following the completion of Stage 1 construction works (described below). Some additional cracking was observed along the crest, particularly on the upstream face, but no damage to the Stage 1 construction works was observed.
4. Dam Deficiencies
A number of deficiencies were identified during the forensic investigations which contributed to the poor performance of the dam, both on first filling and during subsequent seismic events:
The core material, which was constructed from mudstone, was compacted without sufficient preconditioning to bring the mudstone to within the specified gradation. Furthermore, the plant used to compact the material was inappropriate for the task, both in its ability to break down the mudstone clods and to achieve specified compaction, and the resulting material comprised a gap-graded silty gravel or gravelly silt.
- The matrix of silt dominating the engineering performance of the core material was compacted at low density with high air voids. When the reservoir rose, the air voids were filled with water upon saturation, resulting in high moisture contents. High moisture correlates with low strength in the fine grained materials. The laboratory testing completed demonstrated that the in situ material strengths were reduced by up to 40% from those assumed by the designer.
- Analysis of field density tests undertaken during construction confirms that the core material was susceptible to collapse upon wetting. The saturation of the soil would have overcome capillary action holding the soil particles stable as well as disaggregating some of the particles by slaking. Crest settlement monitoring shows large settlements of the embankment both before (150-300 mm) and after the earthquakes (up to a further 130 mm). The settlements were much larger than estimated by the designer and much larger than expected for a dam of this type and size.
- Due to the lack of construction and as-built records, the presence of the stabilising buttress at the upstream toe of the shoulder could not be confirmed.
- Analysis of the upstream shoulder failure using the in situ strength results from the recovered core materials indicated a yield acceleration for the critical failure surface of approximately one half of that estimated by the designer. This lower yield acceleration results in the estimated post-seismic deformations under the Operating Basis Earthquake (OBE) and Maximum Design Earthquake (MDE) almost doubling. The observed embankment deformation following the July 2013 earthquake was even greater than estimated during the forensic analysis
- Grading tests completed on a sample from the existing chimney filter indicated that the filter material used in the embankment was non-compliant with the specification (50% of the material was larger than the specified coarse limit).
- Test pits completed on the dam crest indicated that the chimney filter was terminated 1.4 m lower the design height, and 0.4 m below the normal water level of the reservoir. In addition, it was only 0.5 m wide at the crest, rather than the 1.0 m design width.
- A longitudinal crack on the downstream side of the crest, and minor slumping of the upstream shoulder was observed shortly after first-filling.
5. Regulatory Requirements
The dam is classified as a large dam under the New Zealand Building Act (2004) due to the height of the structure (greater than 4 m) and the volume of water retained (greater than 20,000 m3). As such, it required building consent prior to construction of the remedial works commencing. Under the Act, the design criteria are based on the Potential Impact Classification (PIC), in the unlikely event of dam failure. The dam was assessed at the low end of Medium PIC by the original designers, and this assumption was retained for the design of the Stage 1 and Stage 2 works. The New Zealand Building Code (1992) sets out performance criteria for the dam and appurtenant structures, and the New Zealand Society on Large Dams (NZSOLD) Dam Safety Guidelines (2015) sets out design criteria for large dams. Design and construction of the remedial works were completed in accordance with the relevant sections of these overarching documents.
6. Stage 1 Remediation
The remedial works to restore the dam to full operational capacity was completed in two stages. Stage 1 comprised a “Make Safe” solution, with the intent of enabling the dam to continue to operate at the lowered reservoir level (244.35 m RL) indefinitely. The objective of the Stage 1 “Make Safe” design was to design an interim solution to satisfy the relevant dam safety requirements with the reservoir maintained at the lower level, while also allowing the works to be incorporated into the Stage 2 permanent remediation. For practical purposes the solution allowed the spillway to pass a 1 in 1000 year flood as required for a lower end Medium PIC dam.
Figure 7: LLO Pipe Backfill Typical Section
6.2 Stage 1 Works
The key elements of the Stage 1 works are summarised as follows:
- Low-Level Outlet Pipe Installation: Conversion of the existing emergency low level outlet trench into a permanent Low-Level Outlet (LLO) using a DN560 mm Polyethylene (PE) pipe. The pipe was fully encased in concrete upstream of the chimney filter. The upstream end included a flanged connection for the future addition of a control valve during Stage 2. Following installation of the LLO, the trench was backfilled using the embankment cross section to the existing ground level with zoned fill incorporating an impervious central core, alluvial shoulders, and internal drainage (in the form of a chimney filter and downstream blanket drain) to reduce seepage, control pore pressures within the downstream shoulder, and mitigate the risk of piping though the fill. The dam embankment was founded on in-situ Tertiary-age mudstone, with the core keyed into in-situ mudstone. The core consisted of mudstone won from a borrow area upstream from the dam, conditioned to a clay in the borrow area, and compacted to 98% relative compaction within -1% to + 2% of optimum moisture content. The LLO has sufficient capacity to pass the average annual flow in the stream without surcharging and can pass flows up to the 2 year AEP event before the water level in the reservoir exceeds the spillway channel invert level.
- Spillway Upgrade: The existing spillway channel was enlarged slightly to accommodate the 1 in 1000 year flood event (AECOM’s reassessment of this event resulted in a larger design flow for this event than the flow adopted for the original design). The new channel profile comprised a two stage trapezoidal section with the lower section accommodating all but the extreme flood, which will rise into the upper section but remain within the formed channel. Riprap erosion protection was reinstated along the channel, and at the upstream end in the location of the demolished weir and stilling basin.
- Ford Crossing: A ford was constructed across the spillway 50 m downstream from the spillway mouth. It consisted of 4 round culverts encased in concrete. The height of the ford was approximately 1.2 m above the invert of the spillway channel and the culverts are capable of passing up to a 1 in 5 year AEP event with higher flows passing over the ford.
A typical section through the LLO pipe trench backfill is shown in Figure 7, and some photographs showing the installation of the LLO pipe are shown in Figure 8. Stage 1 remedial works were completed over the course of approximately 5 months from March to August 2016. Code Compliance Certification for the Stage 1 works was issued for the works by the regulator, MDC, on 8 November 2016.
Figure 8: Construction of LLO Pipe
7. Stage 2 Remediation
7.1 Design Intent
The intent of the Stage 2 remedial works was to return to dam to full operational capacity with the reservoir at the design Full Supply Level (FSL) of RL 249.5 m, in accordance with relevant dam safety requirements. In order to achieve this, four key remaining deficiencies had to be addressed, namely:
- Dam Crest: Insufficient compaction at the dam crest, and inadequate filter height;
- Upstream Shoulder: Slumping and instability of the upstream shoulder;
- Core: Likely inadequate conditioning of the existing mudstone core;
- Chimney Filter: Non-compliance of the existing chimney filter.
7.2 Design Process
Prior to undertaking detailed design, an optioneering workshop was held. This workshop included the dam owner, the owner’s engineer, the design team, and design reviewer. Discussion at the workshop focussed on remediation objectives, design constraints, design options, and the relative construction costs associated with the options. The central aim of the workshop was to identify a preferred conceptual design solution for the Stage 2 remedial works that was robust, safe, economical, and acceptable to the regulators and the owner.
The workshop identified five remediation options, with one identified as the preferred. This preferred solution was summarised in a technical memorandum and concept drawings, which was issued to the regulator (MDC) and their peer reviewer (MWH) for comment before proceeding to detailed design. The aim of this technical summary was to engage with the regulators early in order to streamline the approval process, allowing an early start to the construction. Following in-principle agreement from the parties concerned, the preferred concept was advanced to detailed design.
7.3 Stage 2 Works
Building consent has been granted for the Stage 2 remedial works, with construction scheduled to commence in November 2017. The key elements of the Stage 2 works are summarised as follows:
- Removal of slumped material from the upstream shoulder: As this was the location of the worst earthquake-induced damage, the observational method will be employed during construction to ensure that sufficient material is removed. This will include the excavation of a number of trial pits, full time observation by the design team, and inspections by the regulator and the regulator’s peer reviewer. This approach is intended to ensure satisfaction from all parties that sufficient material has been removed. This will be followed by rolling of the cut surface to ensure adequate compaction of the remaining in-situ material.
- Construction of an upstream buttress: The buttress is to be constructed from high-strength alluvial gravel material to stabilise the upstream shoulder, with a minimum horizontal width of 4 m. The buttress will reinstate the upstream slope to the original design line, or be 4 m wide, whichever is larger. The material will have a fines content of 10-15%, which will give it the added benefit of acting as a supplementary seepage barrier to augment the existing core. The riprap protection will be reconstructed on the upstream face.
- Reconstruction of the dam crest: Removal and reconstruction of the upper portion of the dam crest. The depth of removed material is subject to inspection and review by the design engineer during excavation, but will be a minimum of 2.3 m deep. To avoid overtopping during the construction works, an adequate freeboard will be maintained above the current spillway invert level. Emergency actions are included in the contract to backfill the crest should an extreme flood event be forecast. Following completion of the upstream shoulder works, the cut surface of the core and shoulders will be proof rolled, and the crest reconstructed, with mudstone core, gravel shoulders and vertical chimney drain, to the design crest level.
- Construction of secondary downstream chimney filter: A portion of the downstream shoulder will be excavated with a 4 m wide bench at approximately RL 239 m, which is about 4 m above the base of the existing downstream central finger drains. The temporary upslope excavation batter will be 1.4H:1V. This temporary slope batter has been defined to ensure a minimum Factor of Safety of 1.3, to ensure that the works can be completed without compromising safety, which ensures maximum cover for the retro-fitted chimney drain within the downstream shoulder. A 4 m deep, 1 m wide vertical chimney drain from the centre of the bench will be constructed and connected to the existing base finger drains in the downstream shoulder. An inclined chimney filter (1 m wide horizontally) will be constructed within the reconstructed downstream shoulder wedge, and a 1 m wide vertical chimney filter will be constructed within the new dam crest. This will provide a continuous chimney filter from the dam crest to the existing finger drains.
- Reconstruction of the spillway weir and road crossing: The spillway weir and road crossing will be re-constructed at the full supply level. This will entail a small embankment with a concrete cut-off, and reinforced concrete ford crossing. The existing stilling basin for the weir will be reinstated immediately downstream of the weir. The weir and stilling basin will be protected by riprap embedded in concrete (stone pitching).
- Upstream inlet valve: A knife valve will be installed on the existing LLO, with a vertical spindle actuator to allow reservoir drawdown in the event of an emergency.
- Construction of a jetty: A jetty will be constructed to provide access to a vertical spindle actuator which in turn is connected to the LLO pipe valve. The jetty fulfils the important role of allowing access to the LLO valve actuator if the reservoir level needs to be lowered urgently following a future large earthquake event.
Figure 9: Stage 2 General Arrangement
Figure 10: Typical Embankment Profile Following Stage 2 Remedial Works
A general arrangement showing the Stage 2 earthworks extent and spillway works is shown in Figure 9 and a typical section through the reconstructed embankment is shown in Figure 10.
7.4 Innovative Remediation
The adopted Stage 2 remediation solution and proposed construction approach includes several examples of flexible design and innovative engineering, discussed in the following paragraphs.
- Upstream buttress material selection: The material nominated for the upstream buttress was primarily selected for its strength characteristics, and has the added benefit of acting as a supplementary seepage barrier to augment the existing core. The gravel source on-site is well-graded alluvial gravel with cobbles, with 10-15% low-medium plasticity clay fines. The gravel constituents provide maximum strength while the fine material gives the material low permeability. In order for the buttress to have sufficient shear strength, the gravel particles must generally be in contact; therefore, a maximum fines content of 15% is required. In order to provide some benefit as a seepage barrier, the minimum fines content is set at 10%. Tight quality control of this material will be employed during construction to ensure compliance with this design intent.
- Downstream chimney filter retrofit: The proposed construction methodology for installation of the retrofitted downstream filter was derived with input from the Stage 1 contractor. This collaborative approach was intended to improve the constructability of the temporary works associated with the cut into the downstream shoulder. This proved to be a worthwhile exercise, with the contractor suggesting several adjustments to the proposed design. A 4 m deep trial trench was excavated in the downstream shoulder during the design phase to confirm that the proposed trench depth could be achieved. This depth needed to be shallow enough to be safe and constructable, yet deep enough to keep earthworks volumes to a minimum. The retrofitted chimney filter, was relatively standard in design, providing filter protection and also mitigating the risks of development of excess pore pressures to the most susceptible parts of the embankment.
- Involvement of the regulator and peer reviewer in the design and construction process: Given the history of non-compliance and earthquake damage at Haldon Dam, it was considered prudent to involve both the regulator and their peer reviewer during both the design and construction phases. This approach was unconventional, but given the uncertainties around the extent of earthquake-induced damage to the existing dam (particularly the upstream shoulder), was seen as the best approach to provide confidence in the design and construction of the works.
The Haldon Dam has had a troubled history, with a number of non-compliant issues during the initial construction, which led directly to poor performance and emergency works to lower the reservoir following the Cook Strait Earthquake sequence in 2013, despite experiencing reasonably modest peak ground accelerations. The adopted two-stage remediation process was intended to give maximum flexibility to the dam owner. The Stage 1 “make-safe” works resulted in a fully compliant dam, albeit with a lowered reservoir, while also forming part of the Stage 2 permanent remediation works. Key features of the design approach included holding a design workshop with the owner and stakeholders prior to undertaking detailed design, careful rationalisation of the upstream shoulder to optimise the competing interests of strength and permeability, contractor and regulator involvement in the design and construction process, and balancing risk and constructability with the chimney filter retrofit. Applying this approach, the objective of achieving a robust, safe, economical design that was acceptable to the regulators and the owner was achieved.
The authors wish to acknowledge the dam owner, Richard Bell of Starborough Creek Holdings Limited, for permission to publish this paper. The AECOM design and construction support team, including Colin Newton, Lewis Thomas, Sally Hayward and Richard Harkness. The regulator, MDC, including Jeff Atkinson, Graham Allum and Guy Boddington. The regulators peer review panel, MWH, including Peter Foster and Matthew Shore. The Stage 1 contractor, Crafar Crouch Limited, including Norm Crafar and Luke Eden.
Fell, R.; MacGregor, P.; Stapledon, D.; Bell, G.; Foster, M. (2015). Geotechnical Engineering of Dams. Second Edition. CRC Press/Balkema.
GNS Science (2013): Report on the seismicity associated with the 21st July (Cook Strait) and 16 August (Lake Grassmere) 2013 earthquakes, prepared as an Affidavit by Neil Alexander Morris of MDC, dated 18 September 2013.
Holden, C. et al (2013). Sources, ground motion and structural response characteristics in Wellington of the 2013 Cook Strait earthquakes. Bulletin NZ Society for Earthquake Engineering 48(4): 188-195.
New Zealand Building Act (2004), including Building Act Amendments (2013).
New Zealand Building Code (1992).
NZ Geotechnical Society (2005). Field Description of Soil and Rock; Guideline for the field classification and description of soil and rock for engineering purposes. NZ Geotechnical Society, December 2005.
NZSOLD (2015). New Zealand Dam Safety Guidelines.
Raeside, J.D. (1964). Loess deposits of the South Island, New Zealand, and soils formed on them. NZ Journal of Geology and Geophysics 7: 811-838.
Rattenbury, M., Townsend, D., & Johnston, M. (2006). Geology of the Kaikoura Area. Institute of Geological & Nuclear Sciences 1:250,000 geological map. Lower Hutt, New Zealand: GNS Science.
URS (April 2014). Haldon Range Dam Remedial Options Report.
USSR (2011). Materials for Embankment Dams.