On 4 September 2010, a Mw 7.1 earthquake struck 40km west of Christchurch, New Zealand, at 4:35am. There were no fatalities. Damage was significant but not catastrophic. The city appeared to have experienced a major earthquake without a major disaster β proof, some said, of New Zealand's earthquake engineering standards.
Then, on 22 February 2011, a Mw 6.2 aftershock struck directly beneath the city at shallow depth. 185 people died. But the secondary catastrophe β unfolding over the following days and weeks across the eastern suburbs of Christchurch β was unlike anything the city's engineers or insurers had anticipated at scale. Entire residential streets in the suburbs of Bexley, Avonside, and Dallington settled, tilted, and sank into ground that had turned temporarily into a slurry. Sand volcanoes β circular mounds of expelled liquefied sand β appeared across thousands of properties. Sewer and water networks across the eastern city fractured and failed as the ground beneath them moved.
Liquefaction caused losses that rivalled the direct shaking damage in the 2010β11 Canterbury sequence β yet almost no commercial earthquake cat model in use before 2010 captured liquefaction as a separate, modelled component. The Canterbury sequence effectively rewrote the modelling playbook for liquefaction, producing the most extensive post-event geotechnical dataset ever collected for a single liquefaction event β and forcing the cat modelling industry to treat liquefaction not as a minor modifier to shaking loss but as a primary peril in its own right in susceptible geographies.
Liquefaction is the sudden loss of shear strength in saturated, loosely packed sandy or silty soils during earthquake shaking. Under normal conditions, soil derives its strength from friction between soil particles β the particles press against each other under the weight of the material above and the contact forces resist lateral movement. When earthquake shaking rapidly cycles the soil, pore water pressure in the spaces between the particles builds up faster than it can dissipate. As pore pressure rises toward the total stress in the soil column, effective stress approaches zero β the particles lose contact with each other, and the soil mass behaves temporarily as a dense liquid rather than a solid.
The process is transient β it lasts only as long as excess pore pressure is elevated β typically seconds to minutes during and immediately after the earthquake. But the consequences persist indefinitely: buildings that have settled, tilted, or been undermined by scour during the liquefaction phase remain damaged after the soil resolidifies. Infrastructure that fractured during lateral spreading remains broken. Sand ejecta expelled to the surface buries gardens, blocks drains, and contaminates groundwater.
Liquefaction requires three conditions to be present simultaneously:
The liquefied soil layer compacts as pore pressure dissipates after shaking, causing the ground surface to subside. Uniform settlement causes buildings to sink; differential settlement β uneven subsidence across the building footprint β causes tilting, cracking, and distortion of structural frames and building services.
Liquefied soil on gently sloping ground or near free faces (riverbanks, embankments) flows laterally downslope or toward the free face. Horizontal displacements of 1β5m can occur over distances of hundreds of metres. Stretches and ruptures buried infrastructure, tilts buildings, and distorts road pavements.
Excess pore pressure drives liquefied sand and water upward through cracks and openings in the ground surface, forming circular mounds of expelled sand (sand volcanoes) and sheets of water-saturated ejecta across the surface. Primarily cosmetic but indicative of severe liquefaction at depth. Ejecta burial of vegetation, drainage channels, and crawlspace access points causes ongoing maintenance problems.
If the liquefied layer underlies or intersects the building's foundation depth, the bearing capacity of the soil is temporarily lost. Buildings can punch through the liquefied layer, experience extreme differential settlement, or in severe cases tilt to the point of structural failure. Deep pile foundations that penetrate through the liquefied layer into denser material are far more resistant.
Liquefaction hazard modelling differs from all other natural peril modelling in a critical respect: the hazard is not a property of the seismic event alone but of the interaction between the seismic event and the specific soil conditions at each location. The same earthquake produces liquefaction in one part of a city and none 500m away β not because the shaking is different (it may be nearly identical) but because the soil conditions differ at metre-to-tens-of-metres resolution. This makes liquefaction hazard inherently site-specific in a way that no other peril matches.
The foundation of any liquefaction hazard model is a soil susceptibility map β a spatial characterisation of which soils are susceptible to liquefaction based on soil type, density, and water table depth. Susceptibility maps are derived from:
Susceptibility varies with water table depth β the same soil may be Very High susceptibility when the water table is <2m below ground and Low when it is >5m. Seasonal variation in water table is rarely captured in static susceptibility assessments.
Susceptibility tells you which soils can liquefy; the hazard model converts this into the probability that liquefaction actually occurs at a given location under a specific earthquake scenario. The primary analytical framework is the cyclic stress approach β comparing the cyclic shear stress imposed on the soil by the earthquake (Cyclic Stress Ratio, CSR) with the cyclic resistance of the soil (Cyclic Resistance Ratio, CRR). When CSR exceeds CRR, liquefaction is predicted. The ratio CSR/CRR β the Factor of Safety against liquefaction β is the key output: values above 1.0 indicate no liquefaction; values below 1.0 indicate initiation.
In a probabilistic cat model, this calculation is performed for each synthetic earthquake event in the stochastic catalogue against each location's soil profile. The resulting probability of liquefaction initiation at each location, for each event, feeds the vulnerability module. The quality of the soil profile data β primarily from SPT and CPT boreholes β is the dominant source of uncertainty in this calculation. Where borehole data is sparse, susceptibility must be inferred from geological mapping at lower spatial resolution, introducing significant uncertainty in the probability-of-initiation estimate.
The 2010β11 Canterbury sequence provided an unprecedented dataset for liquefaction model calibration. The New Zealand government's Canterbury Earthquake Recovery Authority (CERA) conducted systematic geotechnical investigations across the affected city, collecting thousands of CPT profiles and correlating them with observed liquefaction severity across hundreds of thousands of properties. This dataset β combined with high-resolution aerial photography, LiDAR surveys, and insurance claims data β produced the most comprehensive post-event liquefaction characterisation ever assembled. It directly drove the development of new liquefaction vulnerability functions and improved soil susceptibility mapping methodologies that are now incorporated into commercial cat model platforms.
The February 2011 Mw 6.2 event was smaller in magnitude than the September 2010 Mw 7.1 β yet it caused significantly more liquefaction damage across the city. Three factors explain this. First, location: the 2011 event ruptured directly beneath the city, producing much higher ground motion intensity in the urban area despite its lower magnitude. Second, duration: the cumulative shaking from multiple significant events (September 2010, December 2010, February 2011) had progressively weakened soil structure across the liquefiable eastern suburbs, reducing the cyclic resistance ratio before the fatal event. Third, water table: the February 2011 event occurred in summer, when the Christchurch water table was still elevated following the wet September 2010 period. The interaction of event location, soil pre-weakening from repeated shaking, and water table conditions produced the extreme liquefaction severity that no single-event earthquake model would have predicted.
Liquefaction vulnerability is fundamentally different from shaking vulnerability. Earthquake shaking damages buildings by imposing dynamic inertial loads on the structure β the building's mass resists the acceleration and the resulting stresses damage structural elements. Liquefaction damages buildings by removing or distorting their foundation support β the building is structurally undamaged but its connection to the ground has failed. A building that is highly resistant to shaking (stiff, well-detailed RC frame) may be highly vulnerable to liquefaction (if its foundations are shallow) while a building that is vulnerable to shaking (unreinforced masonry) may survive liquefaction relatively well if its foundation footprint distributes load broadly across the ground surface.
The primary intensity metric for liquefaction damage is ground settlement β the vertical displacement of the ground surface due to compaction of the liquefied layer as pore pressure dissipates. Free-field settlement (what the ground would settle in the absence of any building) provides the baseline; actual building settlement depends on the foundation type and depth. Differential settlement β the difference in settlement across a building's footprint β is the component most directly responsible for structural damage: a building that settles uniformly causes repairable but non-structural damage; a building that settles differentially across its footprint experiences structural distortion.
Based on Canterbury 2011 field data for single-storey timber-frame residential on shallow slab foundations. Damage thresholds vary significantly by foundation type β deep pile foundations show negligible damage at settlement levels that cause total loss for shallow foundations. Differential settlement across the footprint, not mean settlement, drives the damage outcome.
| Variable | High Vulnerability | Low Vulnerability | Why It Matters |
|---|---|---|---|
| Foundation type | HIGH Shallow slab-on-grade, timber piles, pad footings in liquefiable layer | LOW Deep reinforced piles into non-liquefiable stratum, raft foundations on improved ground | The single most important variable β determines whether the building is decoupled from ground movement or participates directly in it |
| Building weight and rigidity | HIGH Heavy, rigid structures on small footprint β concentrate load on small area, punch through liquefied soil | LOW Light, flexible structures on large footprint β distribute load and tolerate differential settlement without structural distress | Load concentration determines whether the foundation exceeds the reduced bearing capacity of the liquefied soil |
| Building position relative to free face | HIGH Within 100m of riverbank, embankment, or retaining wall β in the lateral spreading zone | LOW Flat terrain well away from free faces β limited lateral spreading potential | Lateral spreading imposes horizontal displacements on foundations that vertical settlement does not β particularly damaging for buried services and pile groups |
| Soil improvement | HIGH No ground improvement β natural susceptible deposit | LOW Densified or drained ground β vibro-compaction, stone columns, deep soil mixing | Ground improvement techniques increase the cyclic resistance ratio, raising the shaking threshold required to trigger liquefaction |
| Infrastructure type | HIGH Buried flexible pipes (PVC water mains, clay sewers) β fracture at joints under lateral spreading | LOW Flexible jointed steel pipelines, modern ductile iron β accommodate ground movement without fracture | Buried infrastructure is often more vulnerable than buildings in liquefaction events β the economic cost of infrastructure replacement can exceed building repair costs for severe liquefaction events |
Liquefaction sits in an unusual position in the insurance landscape: it is almost universally covered under standard earthquake insurance policies but almost universally excluded from or absent in standard cat model earthquake output. This creates a systematic gap between what insurers have contractually agreed to pay and what their earthquake cat models are estimating β a gap that became dramatically visible in Canterbury 2011 when liquefaction claims substantially exceeded what any insurer's earthquake model had anticipated.
In most markets, liquefaction is covered as a consequence of earthquake under earthquake insurance policies. The standard language covers "loss or damage directly caused by earthquake, including fire, explosion, landslip, and liquefaction resulting therefrom." This means that where earthquake insurance exists and is purchased, liquefaction losses are typically insured β the challenge is not coverage exclusion but model exclusion: the cat model does not estimate the liquefaction component, so the insurer's view of risk and capital allocation does not reflect it.
In New Zealand, the EQC residential earthquake scheme explicitly covers liquefaction as a covered event under land damage provisions. The Canterbury sequence produced the first large-scale EQC liquefaction payout in the scheme's history, and the resulting land remediation programme β involving ground improvement or managed retreat for the most severely affected residential zones β was a multi-billion-dollar government undertaking. The EQC scheme's response to Canterbury permanently shaped how New Zealand's residential earthquake insurance addresses liquefaction risk.
One of the most complex coverage issues arising from liquefaction is land damage β the distinction between damage to the building (covered by building insurance) and damage to the land on which it sits (which may or may not be covered depending on policy wording). In Canterbury, the government's residential red-zoning β compulsorily purchasing severely liquefied properties to avoid the cost of land remediation β blurred the line between land damage and building damage. Insurers found themselves navigating policy language written before large-scale urban liquefaction had been experienced at the scale Canterbury produced. This coverage ambiguity contributed to prolonged claims resolution and a wave of coverage disputes that took years to settle.
| Region | Susceptibility Profile | Insurance Significance | Modelling Maturity |
|---|---|---|---|
| New Zealand (Christchurch) |
Eastern Christchurch sits on Holocene alluvial and deltaic sands from the Waimakariri River system with shallow water tables. Canterbury Plains extensively underlain by susceptible material. 2010β11 sequence produced the most extensively documented urban liquefaction event in history. | Very high β Canterbury losses set the global benchmark for urban liquefaction insured losses. EQC scheme covers residential liquefaction; commercial market exposed above EQC caps. Post-Canterbury land remediation and red-zoning reshaped New Zealand's residential earthquake and liquefaction insurance structure. | Very high β the most mature liquefaction cat model domain globally, driven by Canterbury dataset. CPT borehole data dense across the city. Multiple commercial and academic models available with Canterbury calibration. The Canterbury event is the primary calibration dataset for global liquefaction model development. |
| Japan (Tokyo Bay, Osaka Bay) |
Extensive areas of Holocene deltaic sediment and reclaimed land (landfill) along Tokyo Bay and Osaka Bay shorelines. Water tables shallow across coastal plains. The 2011 TΕhoku earthquake produced widespread liquefaction in reclaimed land areas of Tokyo Bay (Urayasu) despite the earthquake source being 400km away. | Very high β Japan's high-value coastal industrial and commercial infrastructure is substantially located on susceptible ground around Tokyo and Osaka bays. Urayasu liquefaction from 2011 TΕhoku demonstrated that far-field earthquakes can trigger significant liquefaction in well-engineered Japanese cities, producing infrastructure losses comparable to those from shaking. | High β Japan has extensive CPT and SPT borehole data across major urban areas. Commercial earthquake models are increasingly incorporating liquefaction modules for Japan. Urayasu provided a significant calibration dataset. Tokyo metropolitan area liquefaction hazard maps are publicly available at high resolution. |
| United States (San Francisco Bay, Pacific Northwest) |
San Francisco Bay Area has extensive areas of Bay Mud (soft marine clay β not liquefiable) and hydraulic fill (highly liquefiable). The Marina District liquefaction in the 1989 Loma Prieta earthquake is the canonical US liquefaction loss event. Pacific Northwest (Seattle, Portland, Tacoma) has significant liquefiable river delta deposits. | Significant β the San Francisco Marina District event demonstrated urban liquefaction insured loss potential in the US. A Hayward Fault scenario earthquake would produce extensive liquefaction in low-lying Bay Area communities. Seattle's SoDo district and parts of Tacoma on Holocene fill are significant commercial exposure concentrations. | Good β USGS national liquefaction susceptibility maps available. California has detailed county-level susceptibility mapping. HAZUS includes a liquefaction module. Commercial vendor models for California earthquake increasingly include liquefaction components. Pacific Northwest liquefaction modelling improving ahead of Cascadia scenario planning. |
| Turkey (Istanbul, Izmit) |
Significant liquefiable deposits in river delta and coastal areas around the Sea of Marmara. The 1999 Izmit earthquake produced liquefaction across industrial and residential areas near GΓΆlcΓΌk, causing foundation failures in factory and residential buildings alongside the dominant shaking damage. | Growing β Istanbul's coastal industrial and commercial exposure on Marmara Sea shoreline includes liquefiable ground. A repeat North Anatolian Fault event under or near Istanbul β widely anticipated by seismologists β would produce combined shaking and liquefaction losses across the coastal zone. The 2023 KahramanmaraΕ earthquakes produced liquefaction in riverine areas of the affected region. | Moderate β soil susceptibility mapping exists but borehole data is less dense than Japan or California. Commercial earthquake models for Turkey are incorporating liquefaction components incrementally. Izmit 1999 provided calibration data but the dataset is less comprehensive than Canterbury 2011. |
| Southeast Asia (Bangkok, Jakarta, Manila) |
Major Southeast Asian coastal cities sit on thick sequences of Holocene deltaic and marine sediment with very shallow water tables β some of the world's most susceptible soil conditions for liquefaction. Bangkok's Chao Phraya delta, Jakarta's coastal plain, and Manila Bay's reclaimed areas all have very high susceptibility. | Growing but still low insurance penetration. The combination of extreme soil susceptibility and rapidly growing insured asset values in coastal commercial and industrial areas creates a significant emerging exposure. A major earthquake near any of these cities would almost certainly produce severe liquefaction alongside shaking damage, but commercial cat models for the region typically do not include liquefaction components. | Low β limited CPT borehole coverage, sparse post-event liquefaction data. Geological susceptibility mapping exists at national scales but not at the resolution needed for commercial cat model use. An emerging area for model development as insured values in the region grow. |
The most consequential operational limitation: the vast majority of commercial earthquake cat model runs do not include liquefaction by default. Liquefaction modules, where they exist, are typically optional add-ons that must be specifically selected and are available for a limited set of geographies. An analyst who runs the standard Japan earthquake model and presents the output as the portfolio's earthquake loss is producing a number that excludes liquefaction entirely. For portfolios concentrated on liquefiable ground β Tokyo Bay, Christchurch, San Francisco Bay Area, Istanbul coastal zones β this omission can result in a 20β50% underestimate of the total earthquake insured loss at the most extreme return periods.
Accurate liquefaction hazard modelling requires site-specific geotechnical data β CPT or SPT borehole measurements β at spacing of tens to hundreds of metres to resolve the spatial variability of soil susceptibility. This data exists in reasonable density for a handful of well-studied cities (Christchurch, parts of Tokyo, San Francisco) but is extremely sparse for most of the world's earthquake-exposed cities. In the absence of borehole data, models rely on geological susceptibility maps derived from surface geology β coarser and less accurate. The resulting liquefaction hazard estimates carry large epistemic uncertainty that is often not communicated in model output.
Foundation type is the primary determinant of liquefaction vulnerability β more important than construction class or building age. Yet foundation type is not a standard COPE field in most commercial property underwriting workflows. Models are forced to apply generic vulnerability functions that do not differentiate between shallow slab-on-grade and deep pile foundations, producing systematic overestimation of liquefaction loss for deep-founded commercial buildings and potentially underestimating it for poorly-founded residential stock. This data gap is more consequential for liquefaction than for any other peril where foundation type matters.
Standard probabilistic earthquake models treat each synthetic event independently β the occurrence of one earthquake does not change the soil conditions for subsequent events. The Canterbury sequence demonstrated that repeated earthquake shaking progressively weakens susceptible soils, reducing the cyclic resistance ratio over time and making subsequent events more likely to trigger liquefaction at lower ground motion intensities. A soil that is not susceptible to liquefaction under a Mw 6.2 earthquake in isolation may liquefy under a Mw 6.2 aftershock if a prior Mw 7.1 event has already partially disrupted its structure. This sequencing effect is not captured in models that assume soil properties remain constant between synthetic events.
Liquefaction is as damaging to buried infrastructure β water mains, sewer pipes, underground cables, road pavements β as it is to buildings, and in some events the infrastructure repair cost exceeds the building repair cost. Canterbury 2011 required a multi-year NZD 2.5B+ infrastructure remediation programme across the city's eastern suburbs. Standard commercial property cat models estimate building damage; they do not estimate infrastructure loss. For insurers with contingent business interruption coverage linked to utility restoration, or for reinsurers covering municipal infrastructure, the exclusion of liquefaction infrastructure loss from model output understates the total insured cost of a major urban liquefaction event.
The temporary loss of shear strength in saturated, loosely packed sandy or silty soils during earthquake shaking, caused by rapid buildup of excess pore water pressure. The soil behaves temporarily as a dense liquid. Consequences β settlement, lateral spreading, sand ejecta, foundation failure β persist after the soil resolidifies.
The pressure exerted by water filling the void spaces between soil particles. Under normal conditions, soil grains are in contact and carry the overlying load. During liquefaction triggering, cyclic earthquake shaking builds pore pressure until effective stress approaches zero and grain-to-grain contact is lost.
The horizontal displacement of a liquefied soil mass toward a free face β riverbank, embankment, or excavation β during or after earthquake shaking. Can produce horizontal ground displacements of 1β5m over distances of hundreds of metres. The primary mechanism of buried infrastructure damage in liquefaction events.
A circular mound of liquefied sand and water expelled to the ground surface through cracks or openings during liquefaction. A visible surface indicator of severe liquefaction at depth. Canterbury 2011 produced an estimated 400,000 tonnes of sand ejecta across the city's eastern suburbs.
The ratio of earthquake-induced cyclic shear stress to the initial effective vertical stress in the soil β the demand imposed on the soil by the earthquake. Calculated from PGA and soil overburden. Compared against the Cyclic Resistance Ratio to assess liquefaction triggering potential.
The capacity of the soil to resist liquefaction triggering, derived from in-situ geotechnical measurements (CPT or SPT). When CSR exceeds CRR, liquefaction is predicted to initiate. The ratio CRR/CSR β the Factor of Safety against liquefaction β is the primary metric in liquefaction triggering assessment.
A geotechnical in-situ test in which a cone-tipped probe is pushed into the ground at a constant rate, measuring tip resistance and sleeve friction. Provides a continuous profile of soil density and type with depth. The primary tool for liquefaction susceptibility assessment β the Canterbury 2011 dataset is the largest CPT database assembled for a liquefaction event.
A geotechnical borehole test measuring the number of hammer blows required to drive a split-spoon sampler 300mm into the soil (N-value). Used to assess relative density and liquefaction resistance. Less continuous than CPT but provides soil samples. The basis for the original liquefaction triggering correlations developed by Seed and Idriss in the 1970s.
Uneven vertical subsidence across a building's footprint β the most structurally damaging form of liquefaction-induced ground movement. Causes racking of structural frames, distortion of door and window openings, cracking of floor slabs, and fracture of buried services. The primary driver of residential building damage in Canterbury 2011.
Geological sediments deposited in the last 11,700 years β after the last glacial period. Holocene alluvial, deltaic, and aeolian deposits are the most susceptible to liquefaction because they are relatively young (minimally compacted), often fine-grained, and commonly found near modern river systems and coastlines where water tables are shallow.