Kaikōura Earthquake Slope Hazards – Risk Mitigation and Network Resilience

1. Introduction

More than 20 faults are inferred to have moved in the M7.8 November 2016 Kaikōura Earthquake, which resulted in extensive damage to the coastal slopes and caused uplift at several locations. Between Clarence and Oaro, the transport corridor is located on a narrow coastal strip of land between steep mountainous slopes and the Pacific Ocean. Over 80 landslides occurred along this section of the transport corridor, severely affecting State Highway 1 (SH1) and KiwiRail’s Main North Line (MNL). Lengths of around 10 km to the south of Kaikōura and 14 km on the northern section of the coast (Figure 1) were mostheavily affected.

An extensive geotechnical program was commenced by NCTIR in January 2017 to re-establish this nationally vital portion of New Zealand’s transport infrastructure as quickly as practicable. The intent of this paper is to provide a background of the assessment of slope hazard and resulting risk mitigation works carried out as part of the NCTIR recovery program.

Figure 1: Site locations with coastal corridor sections indicated

2. Geological Setting

Basement ‘Greywacke’ of the Pahau Terrane forms the hills along the coast south of Kaikōura and much of the North Kaikōura Coast. The Greywacke typically comprises slightly weathered sandstone and mudstone (argillite), often with a mantle of moderately weathered rock close to the ground surface. The mudstone is typically weak and the sandstone is moderately strong to strong. At the northern end of the project area, the basement rock is Tertiary-age slightly weathered, calcareous siltstone with minor silty limestone interbeds and is typically very weak to weak.

Figure 2: A large scale prehistoric landslide (scarp outlined in yellow) identified in LIDAR DEM

Colluvium overlying the Greywacke is typically a mixture of rock fragments, silt and sand. It is widely distributed over the basement rock throughout the project area, where slope angles are less than 45°. The colluvial mantle is typically 0.5 m to 1 m thick near the ridge tops, and increases in thickness downslope, with a maximum observed thickness of approximately 15 m.

The presence of large, pre-existing landslides on the slopes facing the coast is evident in both published topographic contour maps and in LiDAR data obtained after the earthquake (Figure 2). Rapid coastal erosion was likely occurring as sea level was rising until about 6,000 years before present. This toe erosion is inferred to have had a major destabilising effect on the coastal slopes, initiating the landslides and keeping them active. Tectonic uplift of the coastline in the last 6,000 years (likely associated with earthquakes similar to the Kaikōura earthquake) has stranded the landslides above the eroding effect of the sea and reduced the destabilising effect. Consequently, the large landslides are now much less active than they have been in the past, and their behaviour is driven by climatic and tectonic events rather than coastal erosion.

3. Earthquake Effects and Post-Earthquake Observations

The combination of rapid tectonic uplift, coastal erosion and oversteepening of the slopes to form the road and rail corridor had rendered the slopes vulnerable to instability. This was demonstrated in the Kaikōura earthquake and historically in rainstorm events, such as occurred in Cyclone Alison in 1975 (Bell 1976).

Fault movements during the Kaikōura Earthquake caused disruption and displacement of the road and rail formations and structures at several locations, but the main impacts on road and rail were due to earthquake shaking and associated slope failures.

The majority of the slope failures caused by the Kaikōura Earthquake were evacuative rock and debris avalanches (Figures 3 to 6) that, in some cases, involved the release of large volumes of material (in excess of 50,000 m³ at Ohau Point for example, Figure 3). Between these large failures, smaller volumes of slope materials or rock were released downslope. Some of this debris reached the road/rail corridor, but in many cases this material has halted on well vegetated hillslopes which are around 35° to 40°. In addition, widespread hillside and ridge cracking occurred during the earthquake, without downslope release of material.

Slope hazards affecting the transport corridor were typically classified by the NCTIR geotechnical team in accordance with the Varnes Landslide Classification, updated by Hungr et al (2013). Four principal slope failure styles were typically observed, as follows.

Figure 3: Slip P6, Ohau Point Kaikōura North, buried SH1

Figure 4: Slip P4, Kaikōura North, buried SH1 and displaced the rail track

Figure 5: Rock and debris avalanche at the south portal of Rail Tunnel 13, Kaikōura South

Figure 6: Slip P7A, Kaikōura North, buried road and rail. Emergency works underway

3.1 Large Scale Landslides

Pre-existing degraded landslide headscarps and evacuated slopes are common along the coastline, although they are generally well vegetated and difficult to recognize on the ground. While in some cases co-seismic deformation of several metres appears to have occurred during the Kaikōura Earthquake, typically these prehistoric landslides have shown no evidence of post-tectonic movement. Reactivation is therefore anticipated to only occur co-seismically under very large earthquake events, with displacements expected to be up to several metres.

3.2 Rockfall and Rock Avalanche

Rockfall triggered by strong shaking was widespread throughout the project area (Figure 7). Source areas included rock outcrops and deposits of boulder colluvium. The scale of rocks observed ranged from fist-sized cobbles to large boulders up to several metres across. The associated ground damage effects of earthquake shaking on the hillslopes included denuding of soil slopes, tension cracking and dilation of rock outcrops, all of which have increased the potential for future rockfall onto the transportation corridor.

Figure 7: Rockfall debris at site P2 near Irongate Stream, Kaikōura North

3.3 Shallow Colluvial Landslides

Shallow soil failures typically involved failure of the colluvial mantle and weathered highly fractured upper parts of the rock mass, as indicated in Figure 8. After the earthquake, these failures were typically reactivated by or immediately following periods of extended rainfall.

Figure 8: Shallow soil failure with displaced soil block; Site SR26, Kaikōura South

3.4 Debris Flows

The transportation corridor has a history of debris flows triggered by prolonged or intense rainfall in steep, erosion-prone catchments that have inundated the road and railway (e.g. Bell, 1976).

The November 2016 earthquake triggered many landslides that did not reach the road or rail, or failed into streams flowing towards the coast. In several locations the exposed slip debris and colluvium on the slopes have since been mobilised by high intensity rainfall, forming debris flows of water-laden rock and soil material that have inundated the transportation corridor (Figure 9). The debris fans that have accumulated consist of silty and sandy gravel, with some cobbles and boulders, some deposited in layers several metres in thickness.

Figure 9: Debris Flow at Jacob’s Ladder, Kaikōura North, following Cyclone Gita (February 2018).

4. Implications for the Kaikōura Coast

Overseas experience has shown that large earthquakes not only trigger extensive landsliding but also increase the number and intensity of subsequent rainfall-induced landslides (Lin et al., 2006; Zhang et al., 2014). Experience from the 1999 Chi-Chi earthquake in Taiwan and the 2008 Wenchuan earthquake in China shows that the critical rainfall thresholds for triggering landslides and debris flows decrease significantly following large earthquakes, commonly reducing to between 25% and 75% of the pre-earthquake threshold.

The heightened landslide initiation probability persists for over a period of several years, with a gradual return to the pre-earthquake conditions (e.g. Chen et al., 2013, Hovius et al., 2011). In mountainous terrain, it can take significantly longer to return to pre-earthquake levels of landsliding (Li et al., 2016; Tang et al., 2016). Analysis of the distribution and characteristics of coseismic landslides triggered by the 1929 Murchison and 1968 Inangahua earthquakes identified hillslopes that did not fail in the 1929 earthquake but subsequently failed in the 1968 event as a result of the damage previously caused (Parker et al., 2015).

An example of this overall decrease in stability is ‘Slip 29A’ in the Oaro-Peketa section south of Kaikōura (Figure 10) which failed as a result of heavy rainfall during Cyclones Debbie and Cook in April 2017 from an area with no historical record of instability, and that had not failed in the Kaikōura earthquake.

In addition to the increased probability of apparent ‘first time’ slope failures like Slip 29A, many of the debris flow sites that have developed over the time interval since the Kaikōura earthquake have originated from ground damaged or debris accumulated from evacuative failures high on the slopes.

The inference drawn from all available evidence was that, consistent with documentation from overseas and from the Murchison and Inangahua earthquakes, the slopes along the Kaikōura Coast are now much more sensitive to rainfall and earthquake effects, and can be expected to remain that way for several decades.

Figure 10: Slip 29A, following failure in April 2017.

5. Transport Corridor Recovery in an Earthquake Damaged Environment

5.1 Project Requirements

The Asset Owner outcome requirements for the NCTIR program were to achieve an acceptable level of safety and service for day to day operations. For KiwiRail this meant operational controls remaining in place after 30 November 2018 were compatible with business-as-usual operations, and for the New Zealand Transport Agency (NZTA) this meant improving reliability of the post-earthquake damaged network.

Two Key Outcomes were required from the NCTIR Program:

First Outcome – Risk to Life

Target risk levels were determined separately by both Asset Owners, with ALARP principles applying for risks lower than these levels.

The risk estimation for the State Highway was completed using the New South Wales Roads and Maritime Services (NSW RMS) Guide to Slope Risk Analysis (Version 4, 2014). This tool permits the rapid and robust estimate of the level of risk posed to road users from landsliding (including rockfall and mass instability) affecting the transportation corridor.

Slope risk adjacent to the rail corridor was assessed using the KiwiRail developed points based slope hazard rating (SHR) system (Justice, 2012) as a proxy. Sites with a higher points rating typically pose a higher risk to the network than slopes with low ratings.

Second Outcome – Level of Service

KiwiRail and NZTA specified acceptable duration outages compared to return period, expressed as an Annual Recurrence Interval (ARI), for outages ranging from a few hours for frequent events, to over 120 days outage for events in excess of 100-year annual recurrence interval.

5.2 Outcome 1 – Risk to Life

5.2.1 Management Framework

The Risk Management approached adopted by the NCTIR geotechnical team was based around the underlying principles of Landslide Risk Management approach developed by Australian Geomechanics Society (AGS, 2007). The approach considered the following key areas:

1. Hazard Analysis (hazard and consequence analysis);
2. Risk Estimation;
3. Risk Evaluation and
4. Risk Mitigation.

5.2.2 Hazard Analysis

Engineering Geological assessment, in conjunction with a GNS study of slope vulnerability and run-out distances (Massey et al, 2017), was used to assess the potential for each of the identified failure mechanisms to affect the entire transport corridor. This highlighted the possible consequences of future rainstorm and seismic triggered instability.

Geomorphological information including field observations, LiDAR hillshade assessment and surface difference modelling was used to evaluate the relative likelihood of a significant failure at each site. Evidence of slope movement included observation of tension cracks and scarps in the upper part of the slope, in some cases measured (by monitoring) in response to rainfall or earthquake aftershocks.

Fault traces and ridge rents that experienced movement as a result of the earthquake were observed throughout the project corridor. Movement of ridge rents as a result of earthquake shaking is not necessarily associated with landslide movement. However, where vertical offset has been observed on a pre-existing scarp and the sense of movement is downslope it has been interpreted as landslide deformation. By way of example, the Half Moon Bay landslide complex, Kaikōura North, exhibits a range of faults, ridge rents, tension cracks and landslide scarps (Figures 11 and 12).

Figure 11: Oblique Aerial View and LiDAR grey scale Image of Half Moon Bay Landslide Complex, Kaikōura North (photograph location and direction indicated by blue arrow)

Figure 12: Example of GIS- based field mapping data (December 2016 aerial imagery) of area shown in Figure 11.

Outputs from landslide characterisation provided information on the possible size and extent of future failures and informed the spatial distribution of landsliding across the transportation corridor. This data was incorporated into the risk and resilience analysis.

5.2.3 Risk Estimation and Evaluation

Analysis of Frequency of Slope Failure
Rainfall triggering thresholds for landslides have been developed by Glade et al. (2000) for the Wellington region. Given the similar geological setting, these thresholds were considered generally applicable to the greywacke terrain in the Kaikōura area. The Glade et al. (2000) study used antecedent rainfall (based on the rainfall over the preceding 10 days) as an index of soil moisture compared to daily rainfall to identify threshold conditions for landslide triggering. The probability that triggering rainfall will occur can then be determined from the frequency/magnitude distribution of the local rainfall record compared to slope instability records.

Records of rainfall and slope movement have been kept since completion of the immediate post-disaster clearance of slips on the north and south coasts. Data has been systematically recorded since March 2017. Comparison of the daily rainfall, the antecedent daily rainfall index and the estimated size of failures is shown in the Figure 13 and typically indicates:

• Active landslides are prone to further debris movement in very small rain events, proportional to the antecedent rainfall condition and the amount of rainfall on the day of slope failure.
• Antecedent rainfall has a strong influence on the amount of rain required to trigger further slope movement.
• Few failures are initiated under heavy rainfall with low antecedent rainfall.
• Relatively large scale failures are commonly initiated following the cessation of rain under high antecedent rainfall conditions.
• Landslide debris fans are being mobilised into channelised debris flows in modest rain events and cause on-going problems to the transport corridor.
• Slopes that did not show obvious signs of failure or deformation in the earthquake have the potential to develop into large slope failures (e.g. Slip 29A).

Figure 13: Observed Rainfall Induced Slope Failures, March 2017 onwards

Consequence Analysis

Where slope failure interacts with an ‘element at risk’ (i.e. road or rail user) the consequence of that interaction was identified in accordance with the requirements of the NSW RMS system or KiwiRail’s slope rating system. The assessment of the consequence of failure at any one site typically involved assessment of several variables, including:

• Boulder size (derived from landslide characterization reporting)
• Run-out distribution (estimated from site observations, rockfall modelling calibrated from site observations or derived from Fahrboeschung angle assessment (Massey et al, 2017),
• temporal probability (typical average daily traffic) and
• vulnerability (for road users).

The consequence of the interaction of the hazard and the ‘element at risk’ typically depends upon the magnitude of the hazard (e.g. runout reaches rail and/or road) and the relative level of protection afforded to the ‘user’ by the immediate environment (for example, what type of vehicle is the user travelling in).

5.2.4 Risk Mitigation

At any one site, risk mitigation involved implementation of one or several solutions. In general, mitigation options involved engineered stabilisation and protection works. However, non-engineered works, such as operational controls in response to earthquakes or rainfall, or remote monitoring were considered as part of the suite of mitigations.

Stabilisation and Protection Works

Engineered risk mitigation solutions can be considered as either reducing the likelihood (active mitigations) or consequence (passive) of failure. Active solutions typically involved, but were not limited to:

• Scaling, boulder removal and sluicing;
• Bulk earthworks;
• Anchored rockfall netting (Figure 14).

Passive structures included, but were not limited to:

• Catch-ditch earthworks
• Rockfall and Debris Catch Fences/Attenuators (Figure 15)
• Shallow Landslide Barriers
• Earth Bunds and Hybrid bunds/Fences (Figure 16)
• Transport Corridor realignment (Figure 17)
• Rock and Debris Avalanche Shelters
• Remote monitoring

Figure 14: Rockfall Mesh and Anchor installation, Ohau Point

Figure 15: Rockfall Attenuator, Parititahi Tunnel North Portals, Kaikōura South

Figure 16: Hybrid gabion basket wall and rockfall fence, Kaikōura South

Figure 17: Irongate Bridge and SH1 realignment at Slip P2, Kaikōura North

Operational Controls

While the majority of slope risk is likely to be associated with the areas where failure occurred during the earthquake, the experience from SR29A suggested that there was some likelihood of first-time failure occurring in previously unidentified areas was also possible as well as reactivation of pre-historic landslides. With this in mind, rainfall Trigger Action Response Plans (TARPs) were developed for road and rail, with probabilistic thresholds established based on the rainfall-slope failure relationship that was determined for the Kaikōura area. The probabilistic threshold curves developed reflect various probabilities of exceedance in relation to slope failure in accordance with Glade et al (2000). Based on rainfall, the TARP enables assessment of slope risks for running freight trains, SH1 traffic and for construction activities, hence predicting when closures may be needed (Paterson, 2018).

To enable the Asset Owners to sufficiently plan alternatives and notify customers if a route closure is likely, the TARP model was developed into a forecasting tool. Based on rainfall records and rainfall forecast over the future four days, it has been possible to assess the likelihood of slope failure in advance. The levels shown on Figure 13 reflect trigger points for differing actions as shown by Figure 18.

In addition to the TARP, a series of remote monitoring installations were placed in a number of key locations to provide an early warning of slope instability affecting (primarily) the rail line. These fences typically comprised a series of string potentiometers connected to tripwires supplemented with tiltmeters installed on fence posts. On impact on any fence, an alarm was automatically generated to enable train services to be halted until an on-ground assessment could be completed and any failed material removed, if necessary.

Figure 18: Example of Rainfall TARP for rail. The dark blue dot represents the current day. The lighter blue dots represent forecast conditions over the next three days (in order of increasing lightness).

5.3 Outcome 2 – Level of Service

For the NCTIR project, “Resilience” was defined as follows (Mason et al, 2018):
Resilience = Robustness + Redundancy + Response
The future slope performance assessment was based on the NCTIR geotechnical team’s assessment of the landslide areas, supplemented by geomorphological assessment of the slopes which have not failed. The assessment considered the slope failure history and the types of anticipated slope failures, observations of the time taken to clear debris, and the requirements or logistics for repair and reinstatement. The type and extent of any planned or completed engineered works was also taken into account.

Initiatives that strengthen each of these aspects will therefore contribute to overall improvement of the resilience.

Robustness
Options to improve the robustness of the corridor included realigning the road and/or rail, engineered works to reduce the potential for slope failure, and engineered works to reduce the potential for inundation of the corridor. Specific measures adopted included:

• Seaward realignment of the rail and road corridors away from the hazardous hillslopes;
• Construction of new bridges to allow debris flows to pass beneath the transport corridor;
• Slope stabilisation with rock bolts and mesh;
• Up-slope earthworks and slope re-profiling;
• Installation of engineered rock fall fences and landslide barriers;
• Installation of drainage measures to relieve groundwater pressures or control surface water runoff.

Redundancy
There are few options to improve redundancy for the rail as the MNL is the sole rail route between Canterbury and Marlborough. For the state highway, redundancy is improved by:

• Upgrading the capability of the alternative routes (principally the Lewis Pass route through state highways 63, 65 and 7);
• Identifying areas where alternative routes can be quickly established to avoid potentially damaged assets (for example, alternative bypass routes for vulnerable bridges).

Response

The NCTIR Alliance has provided KiwiRail and the NZ Transport Agency with the ability to react quickly in the event of a hazard event. After the completion of the NCTIR project period, the following measures will enhance organisational response for management and maintenance of the network:

Installation of tripwire fences with remote monitoring sensors to notify when rock falls or slips occur;

Monitoring of unstable slopes with GPS sensors and ground surveys to provide a forewarning of slope failure;

Implementation of streamlined response plans for hazard events – there are numerous slope assets that have been built under the NCTIR works, many of which protect both road and rail. Recognising this, KiwiRail and NZTA are embarking on creating a joint maintenance process to determine how combined issues can be best responded to, such that personnel and plant can be mobilised quickly in the event of failure.

The combination of engineering works to improve the robustness of the transport corridor and an improved response via improved maintenance and management practices means that NZTA’s and KiwiRail’s level of service are expected to be met for relatively frequent future events, but could be exceeded in larger events.

6. Conclusions

The November 2016 Kaikōura earthquake caused widespread damage and severe disruption to the road and rail corridor along the Kaikōura Coast. Residual damage in hill slopes that did not fail co- or post-seismically but have been sufficiently weakened to fail in the next large triggering event means that the damage legacy of the earthquake will persist in the landscape for decades.

Recovery of this nationally vital portion of New Zealand’s transport infrastructure was consequently focused on (a) reducing life safety risk and (b) improving the resilience of the network.

For slope risk reduction, recovery works have included a suite of engineered risk mitigation solutions that can be considered as either reducing the likelihood (active mitigations) or consequences (passive mitigations) of failure. These physical works are supplemented by non-engineered measures, including a response plan to elevated rainfall, which allows both Asset Owners to proactively manage risks to the transport corridor.

7. References

• Australian Geomechanics Society (25007) Landslide Risk Management Guidelines. Australian Geomechanics, Volume 42 No 1 March 2007
• Bell, D.H. (1976). High intensity rainstorms and geological hazards: Cyclone Alison, March 1975, Kaikōura, New Zealand. Bulletin of the International Association of Engineering Geology, 14: 189-200.
• Chen, J.C., Jan, C.D., Huang, W.S. (2013). Characteristics of rainfall triggering of debris flows in the Chenyulan watershed, Taiwan. Natural Hazards and Earth System Sciences 13: 1015-1023.
• Hungr, O., Leroueil, S., and Picarelli, L. (2013). The Varnes classification of landslide types, an update. Landslides; April 2014 Vol 11, Issue 2 pp 167-194.
• Glade, T., Crozier, M., Smith, P. (2000). Applying probability determination to refine landslide-triggering rainfall thresholds using an empirical “Antecedent Daily Rainfall Model”. Pure and Applied Geophysics 157: 1059-1079.
• Hovius, N., Meunier, P., Lin, C.-W., Chen, H., Chen, Y.-G., Dadson, S., Horng, M.-J., Lines, M. (2011). Prolonged seismically induced erosion and the mass balance of a large earthquake. Earth and Planetary Science Letters 304: 347-355
• Justice, R (2012). Development of a Slope Risk Rating System for New Zealand Rail. 11th Geomech Conf, Melbourne, July 2012.
• Li, L., Yao, X., Zhang, Y., Iqbal, J., Chen, J., Zhou, N. (2016). Surface recovery of landslides triggered by 2008 Ms8.0 Wenchuan earthquake (China): a case study in a typical mountainous watershed. Landslides 13: 787-794.
• Lin, C.-W., Liu, S.-H., Lee, S.-Y., Liu, C.-C. (2006). Impacts of the Chi-Chi earthquake on subsequent rainfall-induced landslides in central Taiwan. Engineering Geology 86: 87–101
• Lin, M.L. et al (2008). The effects of earthquake on landslides – a case study of the Chi-Chi earthquake, 1999. In Landslides and Engineered Slopes (Chen et al – eds). Taylor & Francis, London. p193-201
• Mason, D., Justice, R., McMorran, T. (2018).  Assessment of the resilience of the North Canterbury transport network following the 2016 Kaikōura earthquake.  Proc. 2018 IPWEA NZ Conference, Rotorua, 20-22 June 2018.  Paper No. 507a, 8 p
• Massey C.I, Taig T.T, Lukovic B, Rosser B, Ries W. (2017) The North Canterbury Infrastructure Recovery Alliance: Pilot study for assessing landslide hazards along the road and rail corridors GNS Science Consultancy Report 2017/185 DRAFT 5 October 2017
• Parker, R.N., Hancox, G.T., Petley, D.N., Massey, C.I., Densmore, A.L., Rosser, N.J. (2015). Spatial distributions of earthquake-induced landslides and hillslope preconditioning in the northwest South Island, New Zealand. Earth Surface Dynamics 3: 501-525.
• Paterson Z. (2018). Development of an Early Warning System for Rainfall Induced Slope Failure Affecting Railway in a Post-Earthquake Setting. 12th YGPC, Hobart (in prep)
• Tang, C., Van Westen, C.J., Tanyas, H., Jetten, V.G. (2016). Analysing post-earthquake landslide activity using multi-temporal landslide inventories near the epicentral area of the 2008 Wenchuan earthquake. Natural Hazards and Earth System Sciences 16: 2641-2655
• Zhang, S., Zhang, L.M