Marlborough dam construction challenges

Published 02 July 2018
Marlborough dam construction challenges


The New Zealand Building Act 2004 requires dams above 4m high, holding more than 20,000 cum, be engineer designed with construction supervision, whilst complying with best practice under the NZSOLD Dam Safety Guidelines. The design challenge with these dams for small agricultural enterprises is achieving an economic solution as funding is usually constrained.

A case study of a typical Marlborough earth dam, one of two 20m high earth dams commissioned by Yealands Wine Group, is presented herein. Natural materials used in embankment construction included mudstone, alluvial gravel, and loess. The technical challenges during construction included dealing with ubiquitous loess soils, and availability of locally-sourced, suitable filter material.

Limited availability of gravel required use of loess within the embankment shoulder. Prior to construction a field trial was conducted and a method specification was developed to determine the performance of the loess and the extent of conditioning required to achieve an acceptable embankment fill material.

Another challenge was economically producing suitable filter material. The common practice of washing to remove fines was not practicable, so an innovative “winnowing” technique was adopted. The high seismicity of the region also presented some interesting design challenges.

Figure 1: Project Location Map


Yealands Dodson’s Dam #1 (Dam#1) is the first of two earth embankment dams designed by AECOM and constructed between late 2015 and 2017 by Yealands Wine Group, who operate a privately owned vineyard in the Marlborough Region (north east of the South Island) of New Zealand (see Figure 1). The Marlborough summer climate is characteristically warm and dry with typical summer daytime temperatures ranging from 20°C to 26°C and an average summer rainfall of 105mm (Marlborough Research Centre, 2016). Security of irrigation water supply to meet demand during dry periods is essential for wine production in the region as well as being important to business risk mitigation. The Dodson’s dams were developed primarily to improve water security when abstraction from the river is restricted and will be operated as pumped storages due to the low catchment inflows.

Dam #1 is a zoned earth embankment dam approximately 20m high with a 130m crest length, and a storage volume of approximately 190,000m3. Irrigation water is abstracted from the Awatere River via a dedicated pipeline.

The dam wall comprises a low permeability central core with internal drainage (in the form of chimney filter and downstream finger drains) to reduce seepage and control pore pressures within the downstream shoulder. The dam embankment is founded on Tertiary-age mudstone with the core (Zone 1) keyed into in-situ weathered mudstone, as shown on Figure 2. To minimise construction cost due to limited budget, it was essential to maximise the use of on-site materials and minimise processing of filter materials.

The core consists of compacted mudstone won from the reservoir, and the upstream and downstream shoulders are a combination of loess and alluvial gravel (Zone 3A) and silty to sandy gravel (Zone 3B), also won from the reservoir. Filter sand and drainage gravel materials (Zone 2A and Zone 2B) were processed from a nearby off-site farm paddock, and riprap (Zone 4) was obtained from the Awatere River as pit-run gravel. A concrete weir spillway with a rock lined discharge channel was constructed on the right abutment to control overtopping from pump control failure or flood runoff. A pipe through the embankment was installed for irrigation supply and as an emergency low level outlet.

Figure 2: Typical Embankment Cross Section

Marlborough is an area of high seismicity with a number of active faults near the dam site and several recorded large recent earthquakes in the region, which warranted a defensive approach towards earthquake design.

The dam was constructed by direct forces and plant of the Owner. The construction team was relatively inexperienced in zoned earth dam construction and therefore it was important to develop a design that was simple to build. A combination of factors including the ubiquitous loess soils, limitations around filter material processing, and high seismicity created several design and construction challenges on the project.

This paper describes the methodology adopted for resolving the technical and practical challenges on the project from design through construction.

2 DAM Classification and design criteria

Dam #1 is classified as a large dam under the New Zealand Building Act 2004 (Act) due to the height of the structure (greater than 4m) and the volume of water retained (greater than 20,000m3), and consequently the design and construction is required to be certified by a Recognised Engineer under the Act, a Chartered Professional Engineer with dam design experience. The New Zealand Society on Large Dams (NZSOLD) Dam Safety

Guidelines 2015 underpin the Act and provide the technical guidelines adopted by dam designers and Regulators as the industry standard.

The dam was assessed as Low Potential Impact Classification (PIC) under the Guidelines as in the event of failure the assessed Population at Risk (PAR) is most likely be zero and damage to the environment, property, infrastructure and the local community is assessed as low to moderate. The main impact would be the loss of irrigation water storage (i.e. business risk) in the event of a dam breach. In consultation with the client the PIC was determined to be at the low end of Low PIC. The key risks to dam safety were assessed as the local seismicity and the potential for internal erosion of the dispersive loess material used in the shoulders.

The NZSOLD Guidelines for a Low PIC dam stipulate the following design criteria:

  • Inflow Design Flood (IDF): annual exceedance probability (AEP) 1 in 100 years.
  • Wind and Waves Hazard: Freeboard at IDF Level superimposed with a 1 in 100 year wind AEP (with allowance for wind setup and wave run up).
  • Earthquake Hazard: Operating Basis Earthquake (OBE) AEP 1 in 150 years, and Safety Evaluation Earthquake (SEE) AEP f 1 in 500 years to 1 in 1,000 years. Due to high seismicity of the area a design SEE of 1 in 1,000 AEP was adopted.

Probabilistic estimates of seismic hazard using published data (e.g. NZS 1170.5 (2004) and NZTA Bridge Manual (2013)) estimated a horizontal peak ground acceleration (HPGA) of 0.31g for an OBE event and 0.69g for the SEE event.

Table 1: Summary of Dam Construction Material Engineering Properties


The local geology comprises Tertiary aged sandstone and siltstone (Starborough Formation), overlain by Quaternary-age alluvium (Rattenbury et al. 2006), and mantles of loess with thickness of up to several meters. The Quaternary-age alluvium deposits generally consist of poorly to well graded gravel with sand and silt. Loess commonly comprises silt with clay and fine sand.

The dam construction materials were selected from the materials available on site, which were characterised by pre-design site investigations. These investigations confirmed the typical soil profile to include topsoil overlying loess with gravel lenses of variable thickness. The bedrock was found to be mudstone with an upper weathered zone. The mudstone and gravel are accepted locally as suitable dam construction materials. The loess, although recognised as being problematic as a dam construction material due to its moisture sensitivity, could not be overlooked due to the limited availability gravel and the cost prohibitive alternative of importing embankment fill material.

The characteristics of the adopted construction materials are set out in Table 1 below:


The most interesting design and construction challenges described herein involve the use of the dispersive loess material, and the innovative approach to processing of complying filter material that were successfully used in the development of a cost constrained dam, in an area of high seismicity. These are discussed in Sections 4.1 and 4.2 below:

4.1 Loess Challenges

4.1.1 Loess Characteristics

Due to the natural process of wind-blown deposition, loess has an open (honeycomb) structure with a very high void ratio, but is typically low permeability. Loess can stand at near vertical slopes for long periods provided its moisture content remains low. However, upon wetting, it is highly prone to slumping, dispersion, erosion and piping. Laboratory test results showed the loess to be low to medium plasticity (8 ≤ PI ≥ 12), and results from pinhole dispersion tests (Dispersive D1 with Method A) and crumb tests (Grade 4, Strong Reaction) confirmed its dispersive character. Experience has shown that dispersive soils can be used as construction materials, provided that they are properly designed and well compacted. The main concerns around the use of loess as embankment fill included:

  • Piping through embankment foundation
  • Seismic liquefaction
  • Achieving adequate compaction density due to its moisture content and shear strength sensitivity.
4.1.2 Dam Foundation

In situ loess and compacted loess fill when saturated and exposed to sufficient seismic load can be susceptible to liquefaction and hence the loess beneath the upstream shoulder, the core and the lower areas within the bottom of the valley beneath the downstream shoulder was removed to expose the mudstone basement rock. The mudstone core and vertical chimney filter were keyed into the mudstone foundation along the entire length of the crest. The chimney filter was connected to a series of downstream finger drains with a width and spacing that reduced the extent of saturation within the downstream foundation. This included the loess beneath the u/s shoulder, the core, and the lower areas within the bottom of the valley beneath the d/s shoulder. This reduced the potential for liquefaction as well as piping within the downstream foundation.

4.1.3 Shoulder Design

The concerns around the use of loess in embankment fill construction were primarily around piping risk. To mitigate this risk the use of loess was restricted to those zones of the embankment considered less susceptible to piping. These included the upstream shoulder and an internal zone within the downstream shoulder, immediately downstream from the chimney filter and above the finger drains. Another factor considered in the shoulder design was the stability under the design earthquake loads. In the course of the stability assessment an earthworks optimisation process was applied. This involved balancing total embankment earthworks volumes against gravel fill availability. Use of gravel in embankment shoulder construction allowed steeper embankment batter slopes to be adopted which reduced earthworks volumes but limitations on gravel construction material availability meant that the gravel had to be used judiciously.

During analysis it was also found that due to the low permeability of the loess (5E-09 m/s), rapid drawdown could pose upstream shoulder stability issues, so a 200mm thick horizontal filter layer was included in the lower part of the upstream shoulder to dissipate pore pressures.

4.1.4 Loess Compaction

Loess is well-known as a challenging material to construct fill from due to its sensitivity to moisture changes and blocky characteristics. Prior to construction, a field compaction trial was conducted on Zone 1, Zone 3A and Zone 3B materials to develop a method specification to simplify construction and reduce the need for engineering construction monitoring. The trial determined the compaction density performance of the loess, the best disaggregation method for loess clods, and the extent of conditioning required to achieve an acceptable product without over-wetting (and loss of shear strength). In-situ loess had a moisture content approximately 2%-5% drier than the optimum moisture content (OMC) of 14.5% to 15%, and soil clods contained even lower moisture. The standard compaction tests conducted in a soil laboratory on disaggregated loess material showed a tight compaction curve with a sharp slope on the wetter side of OMC, as shown on Figure 3. Over-conditioned loess material results in loss of shear strength, causing heaving (due to excess pore pressures) at which point the target density cannot be met and the material becomes unworkable.

Figure 3: Loess Compaction Curves

The first step in construction was to mechanically breakdown the loess clods by adding moisture to the fill material. The second step was to compact to required density and void ratio. The construction procedures are summarised as follows:

  • The loess was transported by dump truck from the borrow area to the embankment.
  • A bulldozer spread the material across the fill area to a layer thickness of 200mm. During placement, water was spread on the loose fill surface or on the pile in the front of the ‘dozer.
  • A 30 tonne Caterpillar 825 tamping foot static compactor made 4 passes, at which point the soil clods appeared to be effectively crushed and broken into small fragments. Throughout this procedure, water was only added to the material observed as dry. It was important not to over-condition the fill as it normally took 2 to 3 hours to air dry the material back to OMC. Some experience and judgement, especially during early stages of compaction, was required to avoid over-conditioning of loess fill. To assist with this, simple microwave moisture content tests were conducted on site during the compaction trial and early stage of the construction.
  • The conditioned fill material was then compacted with a 26 tonne padfoot vibrating compactor to further break down soil clods and to achieve the required density. Compaction of each layer was performed in increments of 2 or more passes.
  • The minimum required density of Zone 3A material was 1.76 t/m3, which was 98% of maximum dry density. This was achieved during the trial and throughout the dam embankment construction. Nuclear Density Meters (NDM) testing was used as record testing to confirm specification compliance.

The construction and quality testing procedures established from the construction trial were effective and simple for the constructor to apply.

4.2 Filter Material Challenges

The design gradation for Zone 2A material was to have a fines content of no more than 5% following placement and compaction. In New Zealand, typical alluvial sand contains more than 5% fines, so buying a commercially produced washed sand product would have been very expensive. The owner had access to relatively clean beach sand and gravel from an adjacent property, hence there was a strong commercial motivation to try to produce an unwashed natural product that was fit for purpose. Preliminary PSD testing of representative samples of the beach sediments indicated that it was relatively clean (3-6% fines) and that depending on the specific location within the borrow pit either a coarse to medium sand or a fine gravel, run of pit product could be derived. The sand was identified as a potential Zone 2A material for the chimney filter with the gravel as a Zone 2B material for finger drains. The challenge was establishing a methodology for economically manufacturing a compliant filter material on-site.

Upon completion of the site investigation, the unprocessed Zone 2A and Zone 2B beach sediment materials were confirmed to lack fine sand and so not comply with the design grading limits. The solution was to blend the beach sand with a fine grained sand and remove the excessive fines to produce a complying product.

4.2.1 Filter Material Blending Trial

An alluvial (Wairau) sand source comprising fine sand and silt was found within economic haul distance. A series of blending trials were conducted to produce a product within the Zone 2A filter design envelope. Blending of beach sand and Wairau sand with ratio of 4:1 was found to be the best product. The blended material satisfied the non-erosion filter criteria, but resulted in a product with a fines (<0.075mm sieve) content above 5%.

The Zone 2B material was a pit run beach deposit of well-graded sandy gravel material that contained a high percentage of fines, which had to be removed to meet the filter design criteria.

Figure 4: Zone 2A Chimney Filter Material Particle Size Distribution

4.2.2 Fines Removal Option

Washing the filter material on site to remove fines was not practicable therefore the innovative alternative of ‘winnowing’ out the fines was trialled and then adopted. The dam site is located adjacent to Cook Strait, an area notorious for its strong winds.

To manufacture Zone 2A filter, beach sand was blended with Wairau sand at a ratio of 4:1 using a loader to create a stockpile on a windy day. After this, the material was put through a screen and conveyer to remove oversize, then allowed to gravity fall to the processed stockpile. To improve the effectiveness of the winnowing, the material had to be dry, processed during strong wind, and with adequate falling distance. The same process was adopted for processing Zone 2B material to remove fines from the beach gravel.

Grading of unprocessed and processed of Zone 2A material is shown on Figure 4.

The winnowing technique proved to be effective for processing Zone 2A sand, but not so effective for coarser Zone 2B gravel material. This is thought to be due to the Zone 2B fines being more plastic than the Zone 2A material and/or the dryness of the material as any residual moisture in the stockpiled materials caused the fines to adhere to the granular sand/gravel.

4.2.3 Adjustment of Filter Material in Construction

Variation in grading and fines content for the Zone 2B material meant that despite the best efforts of winnowing described above it was difficult to consistently produce a filter material with fines content below 5%.

Grading limits for the Zone 2B material were adjusted to suit, with an amended grading having a maximum fines content of 6%, as shown on Figure 5. In general, a maximum of 5% (in place) fines was targeted to ensure the material was cohesionless, low plasticity and hence unable to support a crack, and free draining. To allay any concerns over the plasticity, sand castle tests were performed, which confirmed that the material (collapse time under 2 minutes) as very unlikely to be able to support a crack (Fell et al., 2015). Constant head permeability testing on a representative sample confirmed that the permeability of the Zone 2B drainage material (average of 1.3 x 10-3 m/s) was more than adequate for the estimated seepage design flow and finger drain design widths.

The blended material produced was therefore accepted by the designers as adequate.

5 Seismic performance of the dam

The dam site is located in the Australian and Pacific plate boundary transition region with the Hikurangi subduction zone to the northeast and continental collision to the southwest. Movement along the plate boundary is dominantly accommodated by the oblique strike-slip, northeast-southwest 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 Mw7 every 300 to 2,000 years. Four active faults are known within 10km of the site: Hog Swamp Fault, a newly discovered un-named fault in Cook Strait, London Hill Fault, and the Awatere Fault. No faults are known to intersect the site.

Figure 5: Zone 2B Drainage Material Particle Size Distribution

5.1 Historical Seismicity

The dam site is located in an area of known high seismic activity, and a number of large historical earthquakes have been recorded in the region as summarised below:

  • Mw7.4 on the Awatere Fault on 16 October 1848.
  • Mw8.2 Wairarapa earthquake on 23 January 1855.
  • Mw7.3 on the Hope Fault (a key component of the Marlborough Fault Zone) on 01 September 1888.
  • Mw 7.3 (Ms 7.8) Murchison Earthquake on 17 June 1929
  • Mw6.5 Cook Strait Earthquake on 21 July 2013, and Mw6.6 Lake Grassmere Earthquake on 16 August 2013.
  • Mw7.8 Kaikoura Earthquake on 14 November 2016.

In the Lake Grassmere Earthquake, a horizontal PGA of up to 0.7g was recorded in Ward, approximately 17km south east of the dam site (Van Dissen et al. 2013); see Figure 6. Both of the 2013 earthquakes caused damage to buildings on both sides of Cook Strait. However, no surface ruptures were observed.

5.2 Kaikoura Earthquake

The epicentre of the Mw7.8 Kaikoura Earthquake on 14 November 2016, was approximately 150km south-west of the dam site, and movement occurred on or near the interface between the Pacific Plate and the Australian Plate involving a combination of strike-slip faulting and dip-slip faulting on multiple faults. Aftershocks occurred as far north as Wellington, about 150 km to the northeast of the main shock (GNS, 2016).

A GPS site at Cape Campbell (approximately 15km from the dam site) recorded the land shifted to the north-northeast by more than 2m, and up vertically by almost 1m (GNS, 2016). HPGAs recorded were up to 0.67g in Seddon and 1.07g in Ward (GNS, 2016), comparable to those recorded in the 2013 Lake Grassmere and Cook Strait earthquakes (see Figure 7).
Since the earthquake, the site area has been continuously affected by aftershocks with earthquake magnitudes from Mw3 to Mw5.

5.3 Dam Performance after Kaikoura Earthquake

The dam embankment was approaching practical completion and the spillway mouth and chute were being constructed when the Kaikoura Earthquake occurred. In addition, a rainfall event of approximately 35mm over approximately 24 hours followed the day after the main earthquake event.

The HPGA of 0.67g recorded at Seddon (GNS, 2016) is close to the SEE design seismic PGA of 0.69g adopted for the dam. A dam safety inspection was undertaken on 17 November 2016 and observed:

Cracks up to about 20mm wide across the downstream face and along the edge of the filter chimney contact. Settlement appeared to be caused by the earthquake shaking. Most cracks showed an orientation aligned parallel to the crest.

A small amount of clean water flowing from the toe drains (the dam had been partially filled prior to the earthquake).

Rip rap placed in the mid-section of spillway channel was dislodged and the downstream spillway channel was eroded by surface run-off in places.

The earthquake damage appeared to be superficial, and the dam did not show any significant signs of distress. Overall, the dam itself withstood the earthquake and performed well given the level of ground shaking experienced during the Kaikoura earthquake and initial aftershocks.

Figure 6: Two Maps Comparing Peak Ground Acceleration Measured for the Cook Strait (left) and Lake Grassmere (right) Earthquakes (GNS, 2013)

6 Construction Monitoring

As the dam is categorised as a large dam, one of the requirements under the Act is for the designer to conduct surveillance and monitoring of construction quality at specified Hold Points throughout the duration of the construction period, with continued surveillance monitoring once the dam is in service.

A construction quality management plan was drafted by the site engineer to assist the constructor. Pre-construction inspections included borrow pit development, compaction trials, and earthfill, filter and drain material production testing. Construction hold point inspections included a foundation inspection before placement of fill, inlet/outlet pipe installation, spillway construction, and instrumentation installation. Quality control on dam compaction was tested with a Nuclear Densometer. The control testing was carried out and documented by site personnel with periodic checks by the design engineer.

Figure 7: Peak Ground Accelerations Recorded for the Kaikoura Earthquake (GNS, 2016).


This technical paper was supported by AECOM New Zealand and Yealands Estate. We especially thank Peter Yealand of Yealands Wine Group for providing his support during the course of the design and construction of Dodson’s Dam #1.

We would also like to show our gratitude to Don Macfarlane of AECOM (New Zealand) for sharing his wisdom and comments that greatly improved the manuscript.


Fell, R.; MacGregor, P.; Stapledon, D.; Bell, G.; Foster, M. (2015). Geotechnical Engineering of Dams. Second Edition. CRC Press/Balkema.
GNS Science Limited. (2013). Quake pushes Blenheim to the east. Retrieved August 12, 2015, from GNS Science:
Institute of Geological and Nuclear Sciences Ltd. (2004). New Zealand Active Faults Database. Retrieved August 10, 2015, from GNS Science:
Marlborough Research Centre. (2016). Weather Press Releases. Retrieved February 1, 2017, from
New Zealand Society of Large Dams (2015). New Zealand Dam Safety Guidelines.
New Zealand Standard 1170.5: (2004) Structural Design Actions – Part 5: Earthquake actions – New Zealand
NZTA Bridge Manual (2013). Section 5: Earthquake resistant design of structures
Rattenbury, M.S.; Townsend, D.B.; Johnston, M.R. (2006). Geology of the Kaikoura Area. Institute of Geological and Nuclear Sciences 1:250 000 Geological Map 13


Leave a Reply

Issue 95
Volume N/A
Version N/A
Tags , ,
ISSN 0111-6851