Modelling earthquake-induced landslide impacts on infrastructure systems in Wellington

Introduction 

Large earthquakes pose significant risk to life and livelihood in seismically active locations across the globe. Whilst it is widely accepted that earthquake events are responsible for significant economic loss and fatalities, the cascade of secondary hazards instigated by seismic events can be equally, if not more, destructive (Robinson et al., 2016a). The United Nation’s Sendai Framework for Disaster Risk Reduction acknowledges this, calling for multi-hazard approaches to disaster risk reduction (UNDRR, 2015). The risk posed by cascading hazards are thought likely to escalate, as the population continues to grow (Boni et al., 2021), infrastructure systems become increasingly interdependent (UNDRR, 2022), and the frequency of severe weather events rises (Das et al., 2022); thus, quantifying the potential impacts of cascading hazards is critical. Earthquake-induced landslides (EQIL) provide a key example of the risk posed by cascading hazards, with potential impacts that extend beyond those of the ground shaking. Whilst ground shaking alone can result in significant structural damage, EQIL frequently disrupt lifeline infrastructure, exacerbating indirect losses of seismic events through supply chain breakdowns, and loss of transport infrastructure and critical services (Bird & Bommer, 2004). Historic earthquake events have shown greater losses in road infrastructure due to landsliding than that of the initial ground shaking (Bird & Bommer, 2004; Robinson et al., 2016b), highlighting the importance of identifying the potential exposure of critical transport infrastructure to EQIL hazard within seismic hazard analysis. 

There are two traditional approaches to seismic hazard analysis (SHA); deterministic (DSHA) and probabilistic (PSHA). Whilst both approaches have their place in the disaster risk management process, neither approach offers a one-size-fits-all approach to SHA (Bommer, 2002). Robinson et al (2018b) noted that these traditional approaches are not well-suited to contingency planning for a range of plausible earthquake events, which is highlighted by the increasing emphasis on cascading hazards within seismic hazard analysis. Although PSHA considers multiple earthquake scenarios, cascading hazard probabilities and subsequent impacts are difficult to accurately incorporate. With DSHA, including cascading hazards for one, or several, earthquake scenarios is markedly easier, but at the expense of considering the full range of plausible impacts. A more recent approach to SHA has been developed to fill this gap in traditional approaches; ensemble modelling utilises a large range of plausible earthquake scenarios to assess the variability of seismic impacts for a given area, combining strengths of both DSHA and PSHA. Ensemble modelling approaches have historically been developed and applied to the analysis of climate change scenarios, weather forecasts, and ground motion models; however, the use of ensemble modelling for SHA is a relatively recent development, with Robinson et al. (2018b) first employing this method to model fatalities for a number of plausible earthquake scenarios in Nepal. The strength of ensemble modelling is in its consideration of impacts across numerous earthquake scenarios (like with PSHA), with the ability to focus on potential losses from specific scenarios (like with DSHA). Shifting the focus from the likelihood of hazards to the likelihood of impacts for a selected earthquake ensemble allows for the identification of impacts which are consistent across a range of scenarios, as well as those which are scenario-specific, offering insight to the level of certainty of expected impacts, irrespective of the earthquake event that occurs.

The Greater Wellington Region (GWR) has some of the highest seismic risk in Aotearoa New Zealand (Rhoades et al., 2004). Proximity to numerous major faults poses a sizeable threat to metropolitan Wellington (Cousins, 2013), with a rupture of the Wellington Fault considered to be one of the greatest seismic risks to Aotearoa New Zealand (Benites et al., 2003). The extent of this risk is not unknown; the It’s Our Fault programme was established in 2006, to quantify the likelihood, size, effects and impacts of seismic risk in the GWR (Van Dissen et al., 2010). Findings of this programme identified the vulnerability of lifeline infrastructure in Wellington (Barnes et al., 2008), with a potential loss of water, food supplies and transport networks predicted to last for months following an earthquake on any of the major, active faults near the city (Cousins, 2013). Despite the known impacts of EQIL, and the acknowledgement of EQIL risk in existing literature (Barnes et al., 2008; Sadashiva et al., 2021), the potential consequences of this cascading hazard have not been quantified at a high resolution. Further, most DSHA for GWR has focused on a limited number of potential scenarios involving rupture of the Wellington Fault or the Hikurangi Subduction Zone. However, given the number of plausible earthquake scenarios affecting the GWR, as well as the gaps in current understanding of cascading hazard impacts, scenario ensemble modelling presents a potential opportunity for advancing a holistic understanding of seismic risk in this region.

Research objectives and methods 

This study assesses the variability in EQIL hazard and resulting impacts in the GWR for a plausible earthquake scenario ensemble. The earthquake scenario ensemble comprises ten plausible, major (> M_W 7.0) earthquakes scenarios which were selected from the synthetic earthquake catalogue developed by Howell et al. (2023) (Table 1). EQIL hazard was modelled using a fuzzy logic model derived from that of Kritikos et al. (2015) whereby fuzzy membership functions of slope angle, slope position, distance from active faults, and distance from streams, alongside the fuzzy membership of ground motion data for each of the ten selected earthquake scenarios, were established. These functions were aggregated to produce an output (between 0 and 1) identifying EQIL hazard across the research area for each rupture scenario. Outputs of the selected model are not predictive, as there is no temporal component suggesting the likelihood of EQIL occurrence. Instead, the focus is on comparative modelling of the susceptibility of EQIL, referred to as hazard values herein, for a specific pixel under given shaking conditions. 

Exposure of the road network to EQIL hazard was examined for each scenario, assessing the extent of landslide runout reaching the road network and the variability in exposure values of various road classes across the GWR. The landslide runout model employed herein follows that of Robinson et al. (2018a), using a minimum reach angle of 25° as the threshold for landslide runout impacting the road network. This model uses a one-kilometre buffer around points along the road network to identify the source of plausible landslides, restricting the maximum mobility of runout. Within this buffer, pixels with a sufficient reach angle to the road network were identified. Exposure is suggested when pixels with high hazard values, within the established buffer, have a sufficient reach angle for runout that reaches the road network. From the exposure outputs of the road network,  disruption values were derived for key routes across the GWR and access routes for emergency water stations. Six key routes linking Wellington City to the surrounding territorial authorities were examined, with each route commencing at the Wellington Regional Emergency Management Office (WREMO). Access to emergency water was derived from the 21 potential Community Water Stations across Wellington City, Lower Hutt, Upper Hutt and Porirua Cities that were previously identified by Wellington Water.

Key Findings 

The results of this research show the criticality of multi-hazard, multi-scenario approaches for assessing seismic risk; although the extent and severity of EQIL hazard potential was variable across the ensemble, plausible EQIL impacts were present within the GWR across every modelled scenario. Numerous low-specificity impacts have been identified across the GWR; as a result, some impacts can be prepared for even when the next major earthquake is unknown. The outputs of the fuzzy logic model describe EQIL hazard across the GWR for each of the selected earthquake scenarios, with the highest hazard values (0.97) observed in earthquake scenarios on the Hikurangi subduction zone (M_W 7.8), Ohariu (M_W 7.5), Wellington (M_W 7.4), and Wairarapa (M_W 7.6) Faults. Consistently high relative hazard values are observed in Wellington City and Lower Hutt for each of the selected earthquake scenarios, with additional high hazard values across Porirua City, Upper Hutt and the Kāpiti Coast in earthquake scenarios on the Hikurangi subduction zone, and the Ohariu, Wellington and Wairarapa Faults. In the instance of this particular earthquake scenario ensemble, earthquakes of a greater magnitude have not necessarily resulted in greater EQIL hazard, as the selected earthquake scenarios of various magnitudes on the same fault (e.g. M_W 7.6, 7.8 and 8.1 on the Hikurangi subduction zone) do not have the same rupture length or epicentre.

Plausible, yet variable, exposure of the road network EQIL was observed across the earthquake scenario ensemble, with five scenarios giving rise to exposure values ≥ 0.8 along the road network. These stretches of road represent a total length of 0.8 km for a M_W 7.6 and 2.8 km for a M_W 7.8 earthquake on the Hikurangi subduction zone, 11.9 km for a M_W 7.5 event on the Ohariu Fault, 22.8 km for a M_W 7.4 event on the Wellington Fault, and 1.9 km for a M_W 7.6 event on the Wairarapa Fault. Mean exposure of the road network, calculated according to the mean of the exposure values produced by each of the 10 selected earthquake scenarios, gave rise to values ≥ 0.7 across 22 km of road, situated primarily in Wellington City and Lower Hutt. Lower exposure values are observed across the Wairarapa and the Kāpiti Coast. Road segments with high mean exposure values across the ensemble include SH 2 from the Ngauranga Interchange to Petone and the road network along the Wainuiomata coastline from Māhina Bay to Days Bay. These stretches of road have mean exposure values ≥ 0.7, indicating high exposure values across the 10 selected earthquake scenarios. Maximum exposure values ≥ 0.7 can be observed from Paekākāriki to Pukerua Bay in Porirua, through the Akatarawa Valley, and along SH 58 and adjacent roads between Porirua and Upper Hutt; however, the extent of exposure values along these road segments varied across the earthquake scenario ensemble. 

EQIL exposure of the road network acts as a proxy for relative likelihood of disruption to key routes linking Wellington CBD to the other territorial authorities, with similar models utilising hazard values ≥ 0.7 as a threshold for the suggested onset of major disruption (Lin, 2022; Robinson et al., 2018a). Mean exposure values along the most direct route from WREMO to the selected locations of interest range from 0 to 0.74 across the six key routes, with the highest potential disruption located between Ngauranga and Petone on SH 2 (0.73), and Māhina Bay and Days Bay (0.74) in Lower Hutt. High exposure values are also situated on stretches of road between Ngauranga and Johnsonville on SH 1, and along the coast from Seaview to Eastbourne in Lower Hutt. Five of the six key routes have exposure values ≥ 0.7 resulting from numerous earthquake scenarios, suggesting the onset of major disruption along these routes. The potential disruption of SH 2 from Ngauranga to Petone is emphasised, with this stretch of road serving as the highest exposure output in four of the six selected routes: from WREMO to Lower Hutt, Upper Hutt, Eastbourne and Wainuiomata.

The matrix depicted in Table 2 summarises the highest exposure values along the selected key routes for each of the 10 scenarios. High relative disruption is plausible along routes from WREMO to Lower Hutt, Upper Hutt, and Wainuiomata in four of the 10 earthquake scenarios, with potential high disruption from WREMO to Eastbourne in five of the 10 scenarios. Low disruption is expected along the route from WREMO to WLG International Airport, irrespective of the earthquake scenario, and the route to Porirua has potentially moderate disruption in four of the 10 scenarios and low disruption in the remaining six. 

Access to the 21 Community Water Stations (CWS) across Wellington, Lower Hutt, Upper Hutt, and Porirua Cities is considered according to the percentage of road with exposure values ≥ 0.7 within a two-kilometre circular buffer of each CWS, excluding the inner one-kilometre. Exclusion of the inner buffer reflects the walkable distance defined by Mowll et al. (2024), where the distance from a dwelling to the nearest water source is within the distance recommended by the World Health Organisation. The CWS at Days Bay, in Lower Hutt, shows the highest level of potential access disruption, with high disruption in five of the 10 earthquake scenarios; in four of these scenarios, more than 80% of total accessways have exposure values ≥ 0.7. Two of the CWS in Wellington City also have plausible high disruption in the instance of a M_W 7.5 rupture of the Ohariu Fault and a M_W 7.4 rupture of the Wellington Fault; Huntleigh Park CWS and Karori CWS have 78 and 62% of total accessways with exposure values ≥ 0.7, respectively. Of the 21 CWS, 10 are expected to have moderate or high disruption in at least one earthquake scenario within the ensemble. Figure 1 depicts the number of earthquake events which give rise to disruption values greater than zero for each of the 21 CWS, as well as the variation in non-zero values according to the standard deviation. Eight CWS have 80% non-zero values across the earthquake scenario ensemble; however, the standard deviation of these non-zero values ranges from 39.57 (Days Bay CWS) to 7.83 (Aro Valley CWS), thus the predictability of the level of disruption expected from these non-zero values differs greatly between CWS.

Conclusions

This research has demonstrated the value of using a multi-hazard, multi-scenario approach for assessing the variability of seismic impacts. Plausible EQIL impacts were observed within the GWR across every scenario in the selected ensemble, but the location and severity of these impacts were variable according to the seismic scenario. Outputs of this research suggested a number of consistent impacts across the GWR, which were evident in each of the modelled scenarios, including high EQIL hazard in Wellington City and Lower Hutt, plausible EQIL exposure along SH 1 and 2, and on coastal routes along the Wainuiomata Coast, and severe disruption of access routes surrounding the Days Bay CWS. These low-specificity impacts have a high likelihood of occurrence in numerous earthquake scenarios, suggesting these impacts should be expected regardless of the actual next earthquake event. The identified scenario-specific impacts are more uncertain when applying resilience building strategies; whilst these impacts may only occur under specific rupture scenarios, a number of these specific impacts would have significant consequences for the affected areas. 

This high-resolution analysis of EQIL impacts across multiple earthquakes scenarios demonstrates the applicability of a scenario ensemble approach to impact assessment in regions with a number of active faults. Whilst traditional probabilistic and deterministic approaches to seismic hazard analysis have contributed to the existing understanding of seismic hazard in the GWR, utilising an approach that allows for the modelling of impacts from multiple events individually offers a significant contribution to risk and impact assessment in this region. This is a key resource for those tasked with developing and implementing resilience building projects, that presents of a range of plausible impacts, alongside the relative likelihood of these impacts, according to the earthquake scenario. 

This study has also called attention to the variable impacts of cascading hazards in the GWR. The plausibility of EQIL following a major earthquake in the GWR is not a novel contribution; what has been strengthened is an understanding of the variability of cascading hazard impacts, attesting to the relevance of cascading hazards for inclusion in seismic hazard and impact assessments. The criticality of cascading hazards in Aotearoa New Zealand is clear, with landslide impacts resulting from events such as Cyclone Gabrielle and the 2016 Kaikōura earthquake exemplifying the consequences of cascading hazards, and the persistence of their impacts, in the months and years following a hazard event. The necessary inclusion of cascading hazards within seismic hazard research echoes the messaging of the 2015 Sendai Framework; as population growth and infrastructure complexity and interdependence continue to increase, the consideration of multiple hazards, resulting from multiple scenarios, will only become more essential. 

Future work assessing seismic impacts in the GWR, and other seismically active regions, should build on this. The 5,300 earthquake scenarios, ≥ M_W 7.0, suggested in 2022 NSHM lay the foundation for additional applications of scenario ensembles across Aotearoa New Zealand, as do the range of other cascading hazards that have not been considered within the present research. Additionally, inclusion of recurrence intervals in future scenario ensemble model applications would further add to their usefulness for contingency planning efforts, quantifying the probability, and subsequent risk, of expected impacts for a range of earthquake scenarios. These research efforts would enhance the knowledgebase from which existing resilience building projects, such as the 2019 Wellington Lifelines Project, are derived. The complexity of earthquake events cannot be distilled into the magnitude and probability of fault rupture, nor the impacts of ground shaking in isolation. In response, shifting the focus of future seismic hazard and impact assessments away from the likelihood and consequences of singular hazard events is essential in preparing for the next major earthquake.

References 

Barnes, P. M., Pondard, N., Lamarche., G., Mountjoy, J., Van Dissen, R. & Litchfield, N. (2008). It’s Our Fault: Active fault and earthquake sources in Cook Strait. NIWA Client Report WLG2008-56.

Benites, R., Robinson, R., Webb, T. & McGinty, P. (2003). Modelling realistic ruptures on the Wellington fault. GNS Science Report 2003/85.

Bird, J. F. & Bommer, J. J. (2004). Earthquake losses due to ground failure. Engineering Geology, 75, 147-179. DOI 10.1016/j.enggeo.2004.05.006.

Bommer, J. J. (2002). Deterministic vs. probabilistic seismic hazard assessment: an exaggerated and obstructive dichotomy. Journal of Earthquake Engineering, 6(sup001), 43–73.

Boni, M. P., Menoni, S. & Pergalani. F. (2021). Domino and cascading effects in complex events and territorial contexts. Bollettino di Geofisica Teorica ed Applicata, 62(2), 319-334.

Cousins, W. J. (2013). Wellington without water – impacts of large earthquakes. GNS Science Report 2012/30. 124p.

Das, J., Manikanta, V. & Umamahesh, N. V. (2022). Population exposure to compound extreme events in India under different emission and population scenarios. Science of the Total Environment, 806. https://doi.org/10.1016/j.scitotenv.2021.150424.

Howell, A., Penney, C., Kaiser, A. E. & Fry, B. (2023). Modelling ground motions in the Greater Wellington region from multi-fault earthquakes in central Aotearoa New Zealand (GNS Science report; 2023/45). GNS Science. https://doi.org/10.21420/Z9SM-0G27.

Kritikos, T., Robinson, T. R. & Davies, T. R. H. (2015). Regional coseismic landslide hazard assessment without historical landslide inventories: A new approach. JGRES, 120(4).

Lin, A. F. (2022). Seismic exposure and impacts across New Zealand infrastructure networks [Doctoral Thesis, University of Auckland]. UA Research Space.

Mowll, R. M., Anderson, M. J., Logan, T. M., Becker, J. S., Wotherspoon, L. M., Stewart, C., Johnston, D. & Neely, D. (2024). A new mapping tool to visualise critical infrastructure levels of service following a major earthquake. Progress in Disaster Science, 21. https://doi.org/10.1016/j.pdisas.2024.100312.

Rhoades, D. A., Stirling, M. W., Schweig, E. S., & Van Dissen, R. J. (2004). Time-Varying Earthquake Hazard in the Wellington Region. GNS Client Report 2004/14.

Robinson, T. R., Davies, T. R. H., Wilson, T. M. & Orchiston C. (2016a). Coseismic landsliding estimates for an Alpine Fault earthquake and the consequences for erosion of the Southern Alps, New Zealand. Geomorphology, 262, 71-86. http://dx.doi.org/10.1016/j.geomorph.2016.03.033. 

Robinson, T. R., Davies, T. R. H., Wilson, T. M., Orchiston C. & Barth, N. (2016b). Evaluation of coseismic landslide hazard on the proposed Haast-Hollyford Highway, South Island, New Zealand. Georisk: Assessment and Management of Risk for Engineered Systems and Geohazards, 10(2). DOI 10.1080/17499518.2015.1077974.

Robinson, T. R., Rosser, N. J., Davies, T. R. H., Wilson, T. M. & Orchiston C. (2018a). Near-Real-Time Modeling of Landslide Impacts to Inform Rapid Response: An Example from the 2016 Kaikoura, New Zealand, Earthquake. BSSA, 108 (3B). pp. 1665-1682. 

Robinson, T. R., Rosser, N. J., Densmore, A. L., Oven, K. J., Shrestha, S. N. & Guragain, R. (2018b). Use of scenario ensembles for deriving seismic risk. PNAS, 115(41), E9532-E9541. DOI 10.1073/pnas.1807433115. 

Sadashiva, V. K, Mowll, R. L., Uma, S. R., Lin, S. L., Heron, D. W., Horspool, N. A., Nayyerloo, M., Williams, J., Syed, Y., Buxton, R., King, A. B., Lukovic, B., Berryman, K. R. & Daly, M. C. (2021). Improving Wellington Region’s resilience through integrated infrastructure resilience investments. BNZSEE, 54(2).

UNDRR. (2015). Sendai framework for disaster risk reduction 2015–2030. In: UN world conference on disaster risk reduction, Sendai, Japan. Geneva: United Nations Office for Disaster Risk Reduction.

UNDRR. (2022). Global Assessment Report on Disaster Risk Reduction. Our World at Risk: Transforming Governance for a Resilient Future. Summary for Policymakers. UNDRR, Geneva.

Van Dissen, R., Barnes, P., Beavan, J., Cousins, J., Dellow, G., Francois-Holden, C., Fry, B., Langridge, R., Litchfield, N., Little, T., et al. (2010). It’s Our Fault: better defining earthquake risk in Wellington. 11th IAEG Congress. 1 February 2010, Auckland, New Zealand.