1 Introduction

Infrastructure systems, also commonly known as lifelines, can be defined as the systems or networks, which provide for the circulation of peoples, good, services and information upon which health, safety, comfort and economic activity depend (Platt 1991). Infrastructure systems include utility networks such as energy, water, telecommunications, roads and railways or discrete critical facilities such as hospitals, ports and airports. They are crucial to the routine functioning of society and play an important role in emergency relief, reconstruction and recovery after an earthquake.

Despite the seismic fragility of infrastructure systems and the significance of the direct and indirect consequences that loss functionality might have on large communities, the analysis of the earthquake performance and of the post-earthquake management and rebuilding for lifelines appears to be under-represented in the international literature, with respect to buildings and bridges. This is partly due to the main focus being typically given to buildings, which constitute a greater portion of the built environment and pose a greater risk to immediate life safety if seriously damaged. It is also partly due to the paucity, or total lack, of comprehensive data on earthquake-induced damage and loss of functionality of infrastructures, on which to base empirical models or calibrate analytical models.

Both the April 6, 2009 L’Aquila (Italy) moment magnitude (Mw) 6.3 earthquake, and the 2010–2011 Canterbury (New Zealand) earthquake sequence (including amongst others the Mw 7.1 September 4, 2010 Darfield event and the Mw 6.2 February 22, 2011 Christchurch event) have provided an unprecedented opportunity to start filling this gap. Both earthquake sequences occurred in the modern-era of earthquake engineering, with a densely urbanised high industrial and residential urbanisation region close to the source of a severely damaging seismic sequence. In both cases crucial information on the infrastructure systems’ characteristics and on the earthquake-induced physical and functional impact data were collected. Furthermore a large number of ground motion records and ground deformation data were accessible allowing for an accurate understanding of the seismic features of the event.

A shortcoming in the study of damaged infrastructure after an earthquake is the lack of a standardised procedure for data collection and reporting. Whilst reports by the Technical Council on Lifeline Earthquake Engineering, TCLEE, a technical group within the American Society of Civil Engineers, ASCE, provide a thorough description of damages and failure mechanisms observed on field missions (Tang and Cooper 2009; Eidinger and Tang 2012), their level of depth and detail does not support a succinct presentation of damage statistics, which is critical for objective and relative comparison of performances from different earthquakes. One step that can be implemented to improve the post-earthquake assessment of infrastructure systems would be the adoption of a standardised framework, including unanimously recognised taxonomy and impact scales.

Section 2 of this paper presents a non-exhaustive literature review on taxonomies and physical damage descriptors for infrastructure systems developed and used within platforms for assessing seismic risk, including HAZUS (NIBS 2003), Syner-G (Pitilakis et al. 2014a, b) and the Global Earthquake Model, GEM (Pomonis and So 2011). The idea advanced by this paper would be to make direct reference to them, or any future improvement of them, for the post-earthquake assessment of infrastructure systems. Furthermore, it is suggested that, to achieve a holistic and effective post-earthquake assessment for infrastructure systems, “non-physical” performances need to be investigated and reported including: functional impacts; interdependencies issues; resilience attributes. A brief literature review on the aforementioned three aspects is included in the same Section of the paper. Details on the adopted framework within the paper for reporting on the post-earthquake assessment of L’Aquila and Christchurch infrastructure systems are provided.

Sections 3 and 4 report respectively on Canterbury and L’Aquila infrastructures systems, including: gas, water, wastewater, road and electric networks.Footnote 1 A common taxonomy, namely the one proposed within the Syner-G project (Pitilakis et al. 2014a) is adopted, in this paper, to present the assets of the different infrastructure systems. Similarly common damage descriptors, i.e. the ones suggested by HAZUS (NIBS 2003), are used to succinctly present the physical impacts on the different asset components for both L’Aquila and Canterbury infrastructures systems. These implementations seem to confirm that standardised formats for post-earthquake assessment of lifelines infrastructures are viable and feasible.

Section 5 discusses the learning after the L’Aquila and Canterbury earthquakes related to interdependency issues, and factors that have contributed to resilience performances, along with still-opened issues and emerging challenges for enhancing the resilience of lifelines infrastructures to extreme events. Conclusions are drawn in Sect. 6.

2 Towards a standardised framework for post-earthquake assessment of infrastructure systems

2.1 Taxonomies for infrastructure systems

The purpose of classifying infrastructures into taxonomy is to group together elements that can be expected to behave similarly following an earthquake. The classification is also related to the requirements of vulnerability and risk assessment studies of large networks. Whilst buildings are discrete structures which can be grouped together based on certain structural characteristics (e.g. material, geometry, year of construction), infrastructure systems are made up of different components whose differences are based on their function as well as their constructive characteristics. As such classification of infrastructure systems is in two parts; firstly identification of the different components that exist in the system and secondly identification of the typologies of that component. For infrastructure systems it is not just the damage to the individual components that is important, but also how the whole system consequently performs. To ensure that this can be properly evaluated, classification of a single infrastructure system should therefore include all constituent components whose damage condition may affect service provision.

Examples of infrastructure system taxonomies that already exist include guidance from TCLEE (Schiff 1997), the HAZUS methodology taxonomy (NIBS 2003) and the Syner-G taxonomy (Pitilakis et al. 2014a). The first two of these are focused towards infrastructures in the United States, whilst the latter is focused towards Europe. The Global Earthquake Model (GEM) also includes taxonomy for infrastructure systems described in its Consequences Database Guidelines (Pomonis and So 2011). The taxonomies are nevertheless broadly similar and each has its own advantages and disadvantages. The Syner-G taxonomy does not include telecommunications infrastructure and oil delivery infrastructure is treated as the same as gas. However neither telecommunications nor oil delivery are considered in this study, and for the systems that are being considered, the Syner-G taxonomy includes all elements described by HAZUS with some additions, and so is potentially more thorough. The Syner-G taxonomy has been created with European systems, such as those in L’Aquila, in mind and also adequately describes the systems in Christchurch. Therefore the Syner-G taxonomy is used in this study. Table 1 presents the taxonomy of components for the systems of interest in this paper. For the full taxonomy, including component typologies refer to the Syner-G publication edited by Pitilakis et al. (2014a).

Table 1 Syner-G taxonomy for infrastructure systems (adapted from Pitilakis et al. 2014a)
Full size table

2.2 Physical damage descriptors

Standardised discrete and qualitative damage scales represent a common way, within a platform for seismic risk assessment, to quantify physical impacts on infrastructure system components. Different damage scales are used depending on the type of component being assessed. All infrastructures can be represented as an arrangement of nodal components (also known as nodes), and linear components (also known as edges). The nodes are the facilities within the infrastructure system and the edges are the components that connect the facilities and connect end-users to the service.

For nodal structures both the HAZUS and Syner-G platforms use a similar damage scale philosophy in which damage is described according to damage severity. For most components, the damage scale is the same in both platforms, in part because the Syner-G platform adopts many of the HAZUS fragility functions. The most notable difference between the two is in the case of electrical substations, in which the Syner-G measures damage at internal sub-component level. However in earthquake reconnaissance it may not always be possible to enter substations in order to record damage at this scale. The GEM Consequences Database Guidelines (Pomonis and So 2011) also includes proposed damage scales for infrastructure. For nodal structures, GEM proposes a six-level scale with, in comparison with HAZUS, an additional level of very heavy damage or partial collapse. However currently, for most infrastructure elements, vulnerability studies are dominated by the use of HAZUS fragility functions due to the limited number available. Comparisons between observed and predicted damages are at their most informative when using complementary damage scales and so it is proposed to use the HAZUS damage scales for this paper for nodal structures.

For linear components (e.g. pipelines), both HAZUS and Syner-G use the repair rate (repairs per kilometre) to measure damage. Damage scales explicitly for electric power transmission and distribution lines are not defined for HAZUS or Syner-G. However as they are linear elements, in this study a damage rate (faults per kilometre) is used to measure damage. Whilst damage rates can be measured objectively from numerical data, the assignment of damage levels requires some judgment and interpretation. GEM proposes a separate damage scale in addition to repair rate, which is the total length of pipeline affected by earthquake damage, separated by breaks and leaks. This scale has the advantage of being simple and arguably more intuitive to interpret—it is simpler to understand the scale of damage when presented as the quantity (or percentage) of pipeline damaged rather than a rate of damage per unit length. However this measurement is also less informative. For example, if a small number of long pipelines in a network each only have one repair, this will result in a high value for length of pipeline affected. Conversely, if many short lengths of pipeline each have multiple repairs, this will result in a lower value for length of pipeline affected, even though the recovery task may be greater. Consequently this measurement of damage is not necessarily indicative of the actual impact. Whilst both damage scales are useful for understanding the extent of damage, the data sources used in this paper for both locations (O’Rourke et al. 2012; Bray et al. 2013; Esposito et al. 2013) present only repair rates.

One useful aspect of both the HAZUS and GEM damage scales is the separation of breaks (caused mainly by permanent ground deformation) and leaks (caused mainly by ground shaking). This is information which should if possible be recorded given their different functionality impacts—a total break in a pipeline would cut connectivity and cause loss of service whereas a leak might only cause a reduction in service capacity but retain system connectivity. It is not included in this paper however since this data was not recorded.

2.3 Reporting on functional impacts

Functional impacts describe the consequences of earthquake damage on service provision to customers for an infrastructure and therefore measure system rather than component performance. There are three paradigms for measuring system performance: connectivity, serviceability and restoration. Connectivity measurements describe the level of customer connections immediately after an earthquake. Serviceability measurements can be more informative though since they describe not just connection levels, but also what quality of service the connected customers are experiencing. Restoration measures the time taken for services to return to normal pre-earthquake levels. This paper only describes the functional impacts on each system qualitatively, while further analysis to present the data in the form of standardised impact metrics is ongoing. HAZUS does not specifically define connectivity or serviceability metrics for infrastructure systems, but does include restoration curves for individual components. There are also some connectivity and serviceability metrics suggested for the Syner-G platform (Pitilakis et al. 2014b).

2.4 Reporting on interdependency issues

Infrastructure systems do not operate in isolation. Rather there are interdependencies that exist between infrastructures, creating a ‘system-of-systems’ in which the performance of one infrastructure is related to the performance of another. Rinaldi et al. (2001) summarised four main types of interdependent relationship: physical, geographical, cyber and logical. Physical interdependency refers to operational links between two systems in which the state of one system is dependent on the output of another (e.g. water supply system requires electric power to operate pumps). Geographical interdependency refers to an actual physical relationship between two systems such as a pipeline traversing a bridge. Cyber interdependency refers to systems connected by informational links, and logical interdependency refers to all other relationships not categorised by the other three types.

An example of a logical interdependency is the impact that damage to highways and telecoms can have on the ability of other infrastructures to implement repair and restoration works, by restricting the movement of personnel and communications with them. Whereas with other types of interdependency, the relationship between systems exists between physical elements of those systems, in this example the interdependency exists between a physical element (road or telecoms equipment) and a human/organisational element (engineering personnel). Given the potential for widespread damage and urgency in the need to deploy resources for repair and restoration, this type of interdependency is especially important for disaster risk studies and should perhaps be described separately, rather than under the umbrella of ‘logical’ interdependencies. This was addressed by Duenas-Osorio and Kwasinski (2012) in their study on the Maule, Chile earthquake in 2010, by referring to this as a ‘logistical’ interdependency. In this paper, interdependency is only considered qualitatively.

2.5 Reporting on resilience attributes

Bruneau et al. (2003) define four key properties for resilient infrastructures: robustness, redundancy, resourcefulness, and rapidity. They can be respectively defined as: robustness as a system’s ability to withstand stress without a loss of function; redundancy as the substitutability of different elements within the system; resourcefulness as the ability of a system to adapt in order to prevent or reduce disruption of the system, and rapidity as the ability to respond to and mitigate disruption in a timely manner. Bruneau et al. (2003) developed a range of performance criteria, summarised in Table 2, for each one of the resilience principles, in relation to four different performance measures, namely: technical, organisational, social and economic. The resilience properties should be investigated as part of earthquake reconnaissance using specifically-designed survey tool. A great example of that is the tool designed by Mitrani-Reiser et al. (2012) for collecting, as part of post-earthquake reconnaissance, data and information useful for assessing the post-earthquake resilience of hospitals.

Table 2 Performance criteria relating to seismic resilience properties (adapted from Bruneau et al. 2003)
Full size table

2.6 Adopted framework

In the following sub-sections, the assets and the earthquake-induced physical and functional impacts to different infrastructures systems in L’Aquila and in Christchurch are reported according to a common structure, including the description of the: (1) assets; (2) physical impacts; (3) functional impacts. A tabular format is introduced, in particular, for reporting on the level of hazard sustained by each asset component (in terms of both transient and permanent ground deformation) and on the level of earthquake-induced physical impact sustained.

The Syner-G taxonomy is used in this study for classifying the different asset components identified within each of the infrastructure systems. The Syner-G acronyms (as specified in Table 1), are used to group and categorise the components. For each group, details on the numbers of components or km of pipes, are specified along with some further attributes (e.g. material, geometry, constructive details) when available. The values representing the transient and permanent ground deformation hazard have been attributed by GIS analysis (where possible) to determine the specific value at the point of interest. For cases where multiple components have been observed in the same damage state, the corresponding observed hazard is reported as a range rather than individually for each component for conciseness. In some cases, for example where the damage locations are not reported, or where there are a very large number of elements at the same damage level spread across the city, it has not been possible to attribute the specific ground motion value and so the full observed range is presented. The cases where the full range is presented rather than specific values are indicated in the tables.

For all infrastructure components, the damage scale chosen has either been the HAZUS scale or adapted from the HAZUS scale (e.g. bridges). The HAZUS Technical Manual (NIBS 2003) provides definitions for each level of damage, including quantitative measures. Therefore in this study, the appropriate level of damage severity is assigned by comparison of the documented damage observations to the HAZUS definitions. In some cases, where damage reports are non-quantitative, the authors have used their judgment based on qualitative damage descriptions to assign damage level.

3 2010–2011 Canterbury (NZ) earthquake sequence

3.1 Basic facts of the 2010–2011 Canterbury earthquake sequence

The Canterbury earthquake sequence in New Zealand had generated in excess of 10,000 individual seismic events by the end of 2011 but the sequence is notable for five major events, namely September 4, 2010 earthquake (Mw = 7.1), December 26, 2010 swarm (max. Mw = 4.7), February 22, 2011 (Mw = 6.2) and June 13, 2011 (Mw = 6.0), December 23, 2011 (Mw = 5.9) (data available at http://www.geonet.org.nz/).

At 4:35 am (NZST) on 4th September 2010, a Mw 7.1 earthquake and aftershock sequence struck close to the town of Darfield on South Island of New Zealand, 30 km west of Christchurch. Christchurch is the second largest urban centre in New Zealand and the Darfield earthquake was the first large earthquake to strike close to an urban centre in New Zealand since the Hawke’s Bay earthquake of 1931 (Giovinazzi et al. 2011). The fault that ruptured to cause the earthquake was previously unknown but is now identified as the Greendale Fault. There were no fatalities and only two serious injuries (Wood et al. 2010). Accelerograms were recorded from 130 sites across South Island, ten of which had peak horizontal accelerations (PHAs) in the range 0.3 g to 0.82 g. One near-fault record, from Greendale, had a peak vertical acceleration of 1.26 g. 18 records showed peak ground velocities (PGVs) exceeding 0.5 m/s, with three of them exceeding 1 m/s. At 12:51 pm (NZST) on 22nd February 2011, an MW 6.2 earthquake struck the city of Christchurch. The earthquake, caused by an unknown fault, was very shallow at about 5–6 km and the epicentre was just 10 km south-east of the Central Business District (CBD), causing extremely high ground accelerations across the city. The earthquake caused 185 casualties, 8600 injuries and widespread damage to buildings and lifelines. This event occurred while the Canterbury region was still recovering from the Darfield earthquake on 4th September, with many structures suffering from compounded damage. The peak ground acceleration (PGA) at the Christchurch Botanic Gardens in the CBD was 0.5 g in the horizontal direction. The highest horizontal peak ground acceleration recorded was 1.41 g (Bradley and Cubrinovski 2011). All the main events of the earthquake sequence, but in particular the 22nd February earthquake, caused unprecedented levels of liquefaction, throughout the southern and eastern suburbs of Christchurch (Yamada et al. 2011) alongside the Avon River. The liquefaction resulted in settlement, lateral spreading, sand boils, and a large quantity of ejected silt mud and water ponding on the ground surface. The basic facts relating to the two main events in the sequence are shown in Table 3.

Table 3 Basic facts for the Canterbury earthquake sequence main events: 4th September 2010 and 22nd February 2011
Full size table

In relation to observed ground motions, the term ‘full range’ corresponds to the minimum and maximum PGA observations from 20 strong motion stations in the Christchurch area, as documented by Bradley and Hughes (2012). Liquefaction levels are measured using the land damage (LD) categories defined and mapped by Cubrinovski and Taylor (2011): Level 0 for no liquefaction; Level 1 for minor liquefaction; Level 2 for minor to moderate liquefaction; and Level 3 for moderate to severe liquefaction. The mapping of land damage categories was undertaken qualitatively by visual inspection on a drive-through reconnaissance rather than related to measured permanent ground displacement (PGD) values.

3.2 Gas network

3.2.1 Assets

The Liquefied Petroleum Gas (LPG) distribution system in Christchurch is operated by Contact Energy (Rockgas) and it is distributed through a reticulated network, subdivided into 189 separately valved zones that can be manually shut off, at a pressure of about 0.9 bar (90 kPa). The network is supplied from one main feed plant, the Woolston Terminal in the southeast of the city, where the LPG is vaporised and pressurised. This is supplemented by a peaker plant at Harewood in the northwest of the city and three backup feeder plants. Further to the main gas network, in the wider Christchurch area, there are several standalone networks fed from gas cylinders or tanks. The main 0.9 bar reticulated pipe network comprises of approximately 180 km of medium density polyethylene (MDPE) pipes with diameters ranging from 63 to 315 mm and typical wall thicknesses of 6 mm, for the 63 mm pipe, and 14 mm for 160 mm pipe. The depth is typically between 600 and 800 mm. The pipes are welded using electro-fusion fittings and polyethylene PE butt-welding, where a MDPE pipe is melted to another MDPE pipe with an electrical current.

3.2.2 Physical impact

After the 4th September earthquake no damage was reported to plants, storage tanks and pipelines for the Christchurch gas distribution network.

After the 22nd February earthquake, minor leaks were detected in the feed plant at Woolston Terminal (O’Rourke et al. 2012). No damage was observed in either the MDPE distribution pipes or to their welded joints, despite the fact that the pipes traversed zones of severe liquefaction and ground deformation (Giovinazzi et al. 2011). The strength and ductility of MDPE meant that these pipes were able to withstand high levels of tensile stress. A few valve pits had moved relative to the road surface but none of these resulted in damage to the valve and connected pipe (Eidinger and Tang 2012). One service lateral was sheared due to the differential movement of concrete casting around the pipe. The concrete in this case was cast by the customer at the end of the lateral rather than the operator. The observed damage is summarised in Table 4.

Table 4 LPG distribution system-Christchurch: main components and attributes, and hazard and physical impact sustained from 22nd February 2011 earthquake
Full size table

3.2.3 Functional impact

After the 4th September earthquake, a back-up generator at the Harewood peaker failed, however the accident did not create any issue to the functionality of the system, due to the redundancy in the provision of facilities (i.e. Woolston Terminal remained fully operational). Also the power was restored at Harewood later the same day (Eidinger and Tang 2012).

After the 22nd February earthquake, the Christchurch LPG network did not experience any functional impact due to: the minor physical damage sustained by all the components (Table 4); presence of diesel engine back-up generators at the feeder plants that guaranteed their functionality despite the loss of/reduced functionality of the electric power network, (Giovinazzi et al. 2011). However, shortly afterwards, the CBD was isolated for safety reasons at the request of Civil Defence and the four feeders of the system and network valves were shut off in stages. Water supply is necessary in the LPG network for fire protection; because of significant damage to the water distribution system, it became necessary to suspend refuelling of the storage tanks at Woolston Terminal. However redundancy in the provision of facilities meant that gas pressurisation was also possible from the Harewood Plant to the northwest of the city and so system restoration was able to begin without the Woolston Terminal being fully operational (O’Rourke et al. 2012).

3.3 Electric power network

3.3.1 Assets

The transmission of electric power is operated at national level by Transpower, with high voltage lines (220 kV) transmitting power from generating sites to demand centres where the power is further transmitted and distributed by local suppliers. The sub-transmission and distribution of electric power in Christchurch is operated by Orion, who receives power from the national grid at five grid exit points (GXPs) and re-distribute it into the sub-transmission network at either 66 kV or 33 kV or 11 kV. The Orion district/zone substations convert power to 11 kV, if required, and then re-distribute it from the sub-transmission network into the primary distribution network from which further re-distribution and step-down transformation occurs to serve customers. Around 70 % of the lines within the city are buried cables, which can be categorised by their insulation type: predominantly either oil-filled or paper-insulated lead-covered armoured or cross-linked polyethylene (XLPE) or polyvinyl chloride (PVC) (Orion 2009). Of the overhead lines in the city, 97 % are low voltage (400 V). Table 5 summarises all infrastructure managed by Orion in Christchurch and additionally the Transpower GXPs.

Table 5 Christchurch Power Network: main components and attributes
Full size table

3.3.2 Physical impact

Electric power infrastructure generally performed well in the 4th September earthquake. Minor damages to transmission GXPs, included: displaced switches in two GXPs; transformer banks tripped in one GXP; damage to a line in one GXP. Two of Orion’s district substations experienced minor physical damage but remained operational and the impact on Orion’s network was caused primarily by transformers tripping. There were around 30 faults on the 11 kV buried cable network (approximately 4 % of cables affected) but there were no faults on buried 66 kV cables or on overhead lines. Two 66 kV cables were crushed at the Dallington bridge at the transition point between the bridge and the adjoining ground.

In the 22nd February earthquake, Transpower infrastructure was not greatly affected with damage concentrated at the Bromley GXP. Here damage occurred to switchgear in the control building and a voltage transformer in the switchyard. The level of damage at Bromley was low considering the severity of liquefaction in the area. Of the 271 11 kV network substations, only St. Andrew’s Hill suffered damage due to ground shaking while one other suffered damage due to rock fall in Sumner, at the foot of the Port Hills (Eidinger and Tang 2012).

Almost 1000 faults were detected on the buried cable network, including on all 80 11 kV cables serving the Brighton and Dallington suburbs. Most of the cable damage occurred in areas of severe liquefaction and typical examples of the type of damage caused by liquefaction are shown in Fig. 1. The average fault rate for Orion 66 kV cables over the two earthquakes was 0.11 faults per kilometre—since 1994 there have been no faults amongst these cables due to plant failure or third party effects (Orion 2009).

Fig. 1
figure 1

Images of damage to 11 kV cables due to the Canterbury earthquake sequence (Photo credits: Andrew Massie, CPIT)

Full size image

The average fault rate for 11 kV cables over the two earthquakes is 0.24 faults per kilometre. This compares to pre-earthquake observed annual repair rates of less than 0.03 faults per kilometre due to plant failure or third party effects (Orion 2009). The repair rates presented in Table 6, have been interpreted from damages described by Kwasinski et al. (2014) for 66 kV cables and by Kongar et al. (2014) for 11 kV cables.

Table 6 Christchurch Power Network: hazard and physical impact sustained after the 4th September 2010 and the 22nd February 2011 earthquakes
Full size table

3.3.3 Functional impact

Over 160,000 of Orion’s customers (from a total of 190,000) lost power immediately after the 4th September earthquake, although by the end of the day, just over 15,000 customers remained without power. In total, Orion estimate 90 million customer minutes of outages after the 4th September earthquake (Kwasinski et al. 2014). Immediately after the 22nd February earthquake, 80 % of the city did not have access to power, in part because automatic shutdown systems functioned as designed, preventing short circuits and potential earthquake-induced fires. Most of the outages were caused by liquefaction damage to cables (Eidinger and Tang 2012), while above-ground components, including substations and overhead lines performed well. However, 50 % of customers had power restored the same day; 75 % of customers had power restored within 4 days and after 2 weeks, 98 % of customers were receiving power (Giovinazzi et al. 2011). In total, Orion estimate 629 million customer minutes of outages after the 22nd February earthquake (Kwasinski et al. 2014).

As for the transmission grid, to prevent sudden loss of load, which can cause power imbalances and increases in frequency, Transpower initiated after the 22nd February earthquake transfer of power from South Island to North Island, using national demand to balance local over-provision of power and thus limiting the frequency increase in South island (Kwasinski et al. 2014).

3.4 Water and waste water networks

3.4.1 Assets

Christchurch gets its water from wells which can supply up to 150,000 m3 per day by tapping into underlying gravel aquifers. The wells system is supported by 27 pumping stations and 59 storage tanks, seven of which are able to store over 5 million litres. Most of the water pipelines in Christchurch, before the Canterbury earthquakes were of asbestos cement (AC) construction but there were also significant lengths of pipelines constructed with cast iron (CI), polyvinyl chloride (PVC) and modified polyvinyl chloride (MPVC) (Bray et al. 2013). Waste water from the urban Christchurch area gets taken to a single treatment plant at Bromley although there are five additional plants to serve rural areas that are within Christchurch City Council’s (CCC) service area, There are over 100 lift station is the waste water system which are necessary to deliver sewage from low-lying areas around the Avon and Heathcote Rivers to Bromley. The most common types of material for waste water pipelines are reinforced concrete with rubber rings (RCRR), earthenware (EW) and un-plasticised polyvinyl chloride (UPVC), with some concrete (CONC) and AC. The assets within the water and waste water networks are summarised in Tables 7 and 8.

Table 7 Christchurch Water Network: main components and attributes
Full size table
Table 8 Christchurch Wastewater Network: main components and attributes
Full size table

3.4.2 Physical impact

3.4.2.1 Water supply

After the 4th September earthquake, only one storage tank was damaged—in the years preceding the earthquake, Christchurch City Council implemented a seismic upgrade program for all tanks. Eight wells failed and a further one was damaged, all due to liquefaction. One masonry pumping station building was also damaged. Wells and pumping stations that did not have back-up generators lost power immediately after the earthquake, but since some wells were artesian (with around 10 m of head) there was sufficient water available to meet local demand (Eidinger and Tang 2012). Pipes in liquefaction areas (approximately 5–10 % of the city), were severely damaged with approximately 280 repairs required, mostly completed within a week. The failure of pipes and wells result in a loss of pressure in the water system. Fortunately fire was not a major factor after the earthquake so the loss of pressure had no significant consequences.

After the 22nd February earthquake, two storage tanks were damaged, one due to rock fall at McCormack’s Bay and one due to ground failure at Huntsbury. The Huntsbury reservoir was the primary supplier for Christchurch CBD. Since water demand in the CBD after the earthquake was negligible, the loss of this reservoir did not have a critical impact. 20 wells were damaged, in some cases inundating local streets with water (Giovinazzi et al. 2011). The February earthquake resulted in 3000 pipe repairs spread across the city (Eidinger and Tang 2012) and took 6 weeks to complete. It was notable that at Horseshoe Lake and Lyttelton, old asbestos cement and cast iron pipes were replaced by new HDPE pipes due to damage in the September earthquake. No damage was reported in the new HDPE pipes following the February earthquake.

O’Rourke et al. (2012) analysed pipeline repair rates from both earthquakes for different material types. They found that the majority of water supply pipeline repairs occurred in AC pipes although this was unsurprising given the high exposure of this material. In terms of repair rates, AC pipes had the highest rate in liquefaction areas, whilst CI pipes had the highest rate in non-liquefaction areas. In both areas, AC and CI pipes had significantly higher repair rates than PVC or MPVC pipes. Overall, the repair rate in liquefaction areas was 1.68 per km compared to 0.27 per km in non-liquefaction areas. A summary of the damage observed in the water supply system is presented in Table 9.

Table 9 Christchurch Water Network: hazard and physical impact sustained after the 4th September 2010 and the 22nd February 2011 earthquakes
Full size table
3.4.2.2 Waste water

The Bromley waste water treatment plant treats most sewage generated in urban Christchurch up to around 160 million litres per day. The plant experienced some damage due to liquefaction and ground displacement after the 4th September earthquake but this did not affect overall system functionality. There are approximately 1700 km of waste water mains in the Christchurch area, of which about three quarters were made from brittle materials. 200 pipe repairs were required after the 4th September earthquake and two lift stations were damaged by ground displacement (Eidinger and Tang 2012). The pipe damage meant that in some areas, effluent discharged directly into watercourses and there was also a problem of inflow of silts causing the pipes to clog, reducing the overall system capacity.

The 22nd February earthquake caused significant damage to the waste water network and on 2nd April 2011, was deemed to be ‘on the brink of failure’ (Eidinger and Tang 2012), with the main problem being damage to Bromley treatment plant, which was only operating at 30 % of capacity. The treatment plant was also repeatedly damaged by sand and silt, which flowed into broken sewage pipes when the ground liquefied and continuously washed into the basins where the plant was not designed for such heavy solids. This meant some raw sewage had to be disposed of directly into rivers and estuaries. The plant did not return to full treatment processes until July 2011. 6 % of collection pipes were not working and another 27 % were working slowly. Slow moving effluent was a problem and as of 2nd April 2011, the system was leaking 40 million litres per day into backyards and water courses. Groundwater infiltration also exacerbated the demand on the Bromley treatment plant. As well as direct damage due to liquefaction, some pipes that traversed bridges were damaged by the movement of bridge abutments due to lateral spreading. In addition 100 sewer pumping stations were damaged and a number of sewer chambers and manholes experienced flotation.

According to the analysis by O’Rourke et al. (2012), the highest repair rates were found amongst the earthenware and concrete pipes. Overall, repair rates were significantly higher in liquefaction areas than in non-liquefaction areas. For waste water pipes the repair rate was 1.06 per km in the liquefaction zone compared to 0.12 per km outside of this. A summary of the damage observed in the waste water system is presented in Table 10.

Table 10 Christchurch Wastewater Network: hazard and physical impact sustained after the 4th September 2010 and the 22nd February 2011 earthquakes
Full size table

3.4.3 Functional impact

Approximately, 50 % of the city was without water for the first few days following the 22nd February earthquake and more than a third of households were without water for over a week. To increase the supply of fresh water, two temporary desalination plants were set up by the army at Lyttelton and New Brighton and tankers used to distribute water. Within a month over 95 % of occupied units had water, but a “boil order” was in place for over 6 weeks for most of the city due to potential contamination caused by severe damage to the wastewater system. CCC does not normally treat water because of its quality from the aquifers, but as a precaution portable chlorination station were set up across the city and water quality testing was commonplace. The boil orders were only removed once the city experienced 14 consecutive days without water quality issues. Water conservation orders were also put in place as a result of damages to key water reservoirs and the loss of many groundwater pumping wells although most reservoir structures and pump stations performed very well owing to pre-earthquake seismic engineering upgrades (Billings and Charman 2011). Temporary sewage facilities such as chemical and portable toilets were necessary to relieve strain on the waste water system. Though most repairs on the waste water system had been completed by the end of 2011, several hundred homes were still using temporary facilities at that time.

3.5 Road network

3.5.1 Assets

Roads in Christchurch are managed by Christchurch City Council and include over 3,000 km of paved road and 117 bridges. The only tunnel in Christchurch is the Lyttelton Tunnel, connecting the port of Lyttelton to the main Christchurch urban area through the Port Hills. A description of the key assets in the Christchurch road network is presented in Table 11.

Table 11 Christchurch Road Network: main components and attributes
Full size table

3.5.2 Physical impact

Data on road pavement damage due to the February earthquake is taken from the report by Cubrinovski et al. (2014). Road networks were extensively damaged by the significant liquefaction that resulted in settlement, lateral spreading, sand boils and water ponding on the road surfaces. Local roads in the eastern suburbs of the city were the most affected. 83 sections of 57 roads were closed. Five of the six bridges crossing the Lower Avon River were closed and many bridges had weight restrictions implemented. Rock falls in the Port Hills led to several key road closures due to roads being blocked and there was an on-going hazard after the earthquakes from rocks that were in-place but unstable. Closures included Evans Pass, which provided a vital link for oversized or explosive goods between Lyttelton Port and the city, with implications for the LPG delivery. In addition the entire CBD was cordoned off due to collapsed buildings and the perceived risk from standing but damaged structures.

Bridges generally performed satisfactorily in both earthquakes with only minor and moderate damage but with consequences for traffic management and induced damage infrastructure. Many bridges crossing the Avon River suffered tilting in their abutments due to lateral spreading and loss of bearing capacity due to liquefaction (Palermo et al. 2010, 2011). Compared to the Avon River bridges, those crossing the Heathcote River on the southern side of the region were less affected with only a few experiencing minor damage—a typical damage observed was minor settlement of the approach, with little impact on the bridge abutments and superstructure. In general, the damage was most evident between the bridge deck and the approaches. Only the Moorhouse Overpass Bridge suffered significant pier damage during the Christchurch earthquake due to transverse ground shaking (Brando et al. 2012). The relative displacements of bridge superstructures with respect to the bridge abutments and foundations, due to either the ground motion or land-movement, caused leaking/breaks of linkages on water and wastewater pipes and distortion of power (e.g. at Dallington) and telecommunication cables connected to some bridges.

Bridge damage data is presented in Table 12 and has been obtained from the study by Brando et al. (2012), which used a modified HAZUS damage scale in which damage levels D4 and D5 (extensive and complete damage) are combined into a single damage level, D4 for complete damage. It is not possible to attribute these reported ‘complete’ damages back into the original HAZUS damage levels of D4 and D5 and so an exception to the use of HAZUS damage scales is made in this case.

Table 12 Christchurch Road Network: hazard and physical impact sustained after the 4th September 2010 and the 22nd February 2011 earthquakes
Full size table

3.5.3 Functional impact

Most of the State Highways remained open and only one tunnel on the State Highway network, Lyttelton Tunnel, had significant impacts, with rock fall causing damage to the tunnel canopy and control building and loss of power requiring the electrical systems to operate from a back-up generator. However the tunnel re-opened on 26th February, initially for restricted use by residents of Lyttelton, emergency services and fuel deliveries (Lyttelton being the major port for the Canterbury region). Substantial temporary traffic management measures were put in place to manage the residual functionality of the road network, including temporary speed restrictions to 30 kph; adjustments to traffic signals; and adjustments to bus routes. Despite the temporary traffic management measures and the significant programme to hasten the liquefaction clean-up operations, congestion remained problematic for months following the earthquake.

4 L’Aquila (Italy) earthquake

4.1 Basic facts of the 2009 L’Aquila Mw = 6.3. earthquake

On April 6th 2009, 1:32 am (UTC), a Mw 6.3 earthquake struck the Abruzzo region, in central Italy. The earthquake occurred at about 10 km depth along the Paganica fault, located right below the city of L’Aquila (INGV 2009). It was followed by several aftershocks, including 12 (Mw > 4.0) events. 305 people were killed, 1500 injured, about 70,000–80,000 people were temporarily evacuated after the main event. More than 24,000 people were left homeless due to the severe and extensive damage to buildings and lifelines servicing the buildings. The damage to structures and infrastructures was detected over a broad area of approximately 600 square kilometres, including the downtown of L’Aquila and several rural villages. The PGA recorded in the near-source region ranged from 0.33 to 0.65 g, the latter representing one of the highest PGA values measured in Italy (Chioccarelli et al. 2009). With respect to geotechnical effects induced by the earthquake, evidences of surface rupture was found along the Paganica fault including a set of well-aligned ground ruptures that caused significant damage to the crossing infrastructures (Dolce et al. 2009). Further earthquake-induced geotechnical secondary hazards included: slope instability, and numerous rock falls within the Gran Sasso Mountain; collapse of some underground cavities; minor ground settlement induced by liquefaction, and cases of sand boils and sand volcanoes developed in free field in an area, approximately 45 km far from the epicentre (Monaco et al. 2009). The basic facts of the L’Aquila earthquake are presented in Table 13.

Table 13 Basic facts for the 2009 L’Aquila earthquake
Full size table

4.2 Gas network

4.2.1 Assets

The L’Aquila gas network is operated by Enel Rete Gas and distributed via a 621 km pipeline network: 234 km operating at medium pressure (MP) (2.5–3 bar) and the remaining 387 km at low pressure (LP) (0.025–0.035 bar). The MP network is connected to the high-pressure transmission network through three metering and reduction stations (M/R stations) cased in one-storey reinforced concrete structures with steel roofs. Pipelines of medium- and low-pressure distribution networks are either made of steel or high-density polyethylene (HDPE). The latter pipes use fusion joints and have nominal diameters ranging from 32 to 400 mm. For steel pipes gas arc welded joints are used while pipe diameter is usually between 25 and 300 mm. The transformation of the MP into the LP is operated via 300 reduction groups (RGs) that are either buried, sheltered in a metallic kiosk or housed within/close to a building. Several demand nodes, consisting of buried and above ground pipes and accessory elements, allow the supply of natural gas to utilities, from LP network. The assets of the L’Aquila gas network are summarised in Table 14.

Table 14 L’Aquila Gas Distribution Network main components and attributes and hazard and physical impact sustained after the 6th April 2009 earthquake
Full size table

4.2.2 Physical impact

The performance of the impact of the L’Aquila earthquake to the medium- and low-pressure gas distribution system was reasonably satisfactory (Esposito et al. 2013).

As for the performance of above-ground components of the system, no damage was observed to the buildings and the regulator and mechanical equipment of the three M/R stations, while for RGs, one of the principal causes of damage was the collapse of rubble from adjacent buildings. As a consequence, some RGs had to be replaced concurrently with the laying of new pipe. With regard to buried components, namely pipelines, valves, and demand nodes, the largest proportion of repairs (72 %), was localised on the LP distribution network, made of steel pipes (Esposito et al. 2013). The repair and replacement for the steel pipes of LP distribution network were mainly due to breaks or leaks in correspondence of the gas welded joints.

4.2.3 Functional impact

Regarding the functional performance, in the first hours following the earthquake, the priority identified for the management of the gas network was the securing of the network in order to avoid explosions, gas leaks and fires. To this aim, the entire network was shut off via the closure of the three operating M/R stations (Dolce et al. 2009). Furthermore the 300 RGs were closed and in the days following the event, all the gas valves external to each residential property were closed.

4.3 Electric power network

4.3.1 Assets

Terna operates the transmission of electric power at national level, with high voltage lines (220 kV). Enel Distribuzione S.p.A. operated the sub-transmission and distribution of electric power in L’ Aquila, receiving power from the national grid and re-distributing it into the sub-transmission network at either 150 or 60 kV. The sub-transmission and distribution network in L’Aquila include two transformation-distribution substations, medium and low voltage lines and a control centre building.

4.3.2 Physical impact

Electric power network in L’Aquila sustained damage especially of its critical facilities and equipment (Tang and Cooper 2009; Dolce et al. 2009). The two transformation-distribution substations serving the greater L’Aquila suffered damage in bus connections to transformers due to shifting of the un-anchored transformers during the earthquake. In particular, one of the two transformers moved about 14 cm. As for the distribution systems 30 posts were damaged causing service disruption. More than 180 pedestal type connection boxes were dislocated and severed cable connections at the termination lugs that resulted in localised power failure. The electric network control centre building was severely damaged implying a relocation of the control equipment to a temporary shelter next to the building.

4.3.3 Functional impact

In some locations, the electric power was out for a period of 3 days, e.g. wastewater treatment plant (Tang and Cooper 2009) but the essential part of the recovery operation was completed in less than 12 h. The essential part of the system was functional by 9 am the day after the earthquake and about 80 % of the 15,000 customers were reconnected.

4.4 Water and waste water networks

4.4.1 Assets

Gran Sasso Acqua G.S.A. S.p.A provides an integrated water service in L’Aquila city including potable water supply, sewerage and waste water treatment. The water supplied is transported by a network consisting of approximately 900 km of large diameter pipes and is stored in a huge number of tanks (about 200). The water is distributed from the tanks to approximately 100,000 customers through a 110,000 km distribution network made of quite old cast iron and steel pipes (Dolce et al. 2009). The pressure inside the main pipeline network is quite high, reaching 30–50 atm, as well as in the distribution networks where it can reach 6–8 atm. The system is also monitored by a remote control service and guided valves connected through cables or wireless connection, to the main reservoirs and supply systems that allows the check of water flow inside the pipelines, the partial or total management of opening/closing operations and the assessment of the water level in the tanks.

4.4.2 Physical impact

4.4.2.1 Water supply

On the morning of April 6th, 2009, a significant and sudden change in the water flow for a main transmission pipeline was observed by the remote control system. Before the technician team reached the affected site, the relative shutters for that pipe was closed by GSA headquarters. The steel joint of the pipeline (diameter ∅ = 600 mm; pressure 25–30 atm) slipped off, causing a powerful discharge of water from the main pipe. The cause of damage was identified as co-seismic rupture of the Paganica fault that crossed the pipe. With the exception of the aforementioned joint slip, no significant damage was observed to the main distribution and storage system. In the minor distribution system slippage/breakage of the joints and the breaking of cast iron pipes were the most commonly observed damages.

4.4.2.2 Waste water

Regarding the waste water network, treatments plants sustained significant damage (Dolce et al. 2009; Tang and Cooper 2009). The two treatment plants serving the greater L’Aquila area (Ponte Rosarolo and Pile) are located near the epicentre and close to the AQK accelerometric station, which recorded ground accelerations equal to 3.7 m/s2. The older treatment plant (Rosaloro) suffered partial collapse of a longitudinal wall, several vertical cracks on a transversal wall and the separation of orthogonal walls at the edges. The partial collapse of the wall also involved the steel pipe adducting waste water that was connected to it. In buildings used as offices, local technological and laboratory equipment (RC framed structure) were also found cracks of both internal partitions and external walls. At the newer treatment plant in Pile, structural damages were not detected and only some damages to the partitions of local offices occurred. With regard to the older settling tanks, characterised by a circular cross section, a deterioration of the curbing RC beam was detected due to significant corrosion of the steel reinforcements.

At the other plants (Corfinio and Arischia), quite distant from earthquake epicentre, minor structural damages were observed. In particular, at the plant located in Arischia, some cracks on the walls of a distribution trap and damages to the retaining stonewall, were observed. With regard to the circular tanks, one of the two rotating distributors was put out of service for damage to its support. At the Corfinio plant, the central part of the longitudinal walls of the aeration tank showed a rotation probably occurred in large part before the seismic event.

4.4.3 Functional impact

Following the repair of the damaged joint, the provision of potable water for all municipalities administered by the G.S.A. S.p.A. restarted since the evening of April 6th. Regarding the minor distribution systems, since a lot of ruptures were expected in L’Aquila historical centre and in the most affected villages, it was decided to not to restore the water distribution in order to prevent flooding and deterioration in the buildings already damaged. The restoration of the water provision was gradually operated starting from the less affected zones and/or the zones with a strong need for reactivation; this activity was possible because of secondary networks and a shutter system that allowed the exclusion of areas where the water supply was not urgently needed (Dolce et al. 2009).

Regarding the waste water plants serving L’Aquila area, the total capacity was reduced to about 60 %, but at the same time, the demand was also reduced due to the large number of buildings damaged and residents not being allowed to return to their homes (Tang and Cooper 2009).

4.5 Road network

4.5.1 Assets

The road network in L’Aquila city is operated by ANAS S.p.A., that manages in the Abruzzo Region, as well as in the rest of the national territory, the state road network.

4.5.2 Physical impact

The network did not sustain significant damage (Dolce et al. 2009) except for the structural failure of the viaduct “Corfinio” on the national roadway SS5 and the collapse of a bridge on the main road SP36 “Forconese”. No further significant damages were reported to the components of the road networks including the numerous tunnels present in the Region that performed well.

4.5.3 Functional impact

Rock falls and landslide triggered by the earthquake and aggravated by the heavy rain that hit the area in the days following the event affected the network mobility in mountainous areas around L’Aquila. In the urban area, instead, mobility limitations were caused by debris following damaged and/or unsafe residential and monumental buildings adjacent to the roads.

5 Lessons learned

5.1 Technical performance

Although there was some disruption to all infrastructures in Christchurch, the systems demonstrated good levels of robustness. Operators of both the electric power and water supply networks in Christchurch reported that damage to their systems could have been worse but for prior investment in seismic strengthening and retrofitting of facilities. These programs took place despite the fact that at the time, the seismic risk to Christchurch was considered to be low. In the case of the electric power network, Orion implemented their substation strengthening program in response to the Risks and Realities (CAE 1997) lifelines risk project in Christchurch and the surrounding Canterbury region. This forward planning and recognition of the importance of protecting critical infrastructures, even in what were believed then to be low hazard zones, prevented more considerable social and economic consequences.

After the first earthquake Christchurch City Council took quick action in replacing damaged water pipelines with more robust HDPE products. These performed exceptionally well in subsequent events, mitigating some of the impact of the more damaging February 2011 earthquake. Although this work only took place on a small section of the network, this construction embodied the principle of ‘building back better’. Notably HDPE pipelines in the gas network also performed well in the L’Aquila earthquake. Unfortunately the proportion of HDPE pipes as well as the proportion of PVC pipes was very minor in the L’Aquila gas network (mainly made of steel pipes with gas-welded joints) and very absent in the water network, all made of relatively old cast iron and steel pipes. The lack of robustness of such brittle pipe materials and joint solutions in L’Aquila was the main cause of the reduced serviceability of the networks post-earthquake. The disregarding of good practices, such as anchoring of components that could potentially be subjected to out of plane mechanisms under seismic forces, were identified as a vulnerability factor that reduced the robustness of the electric power system in L’Aquila. Transformers in substations were not anchored (and two substations serving the greater L’Aquila had damaged connections between a rigid bus and insulator, due to shifting of the un-anchored transformers), as well as the batteries within the control house of substations that were not anchored or tied to the racks.

The importance of building redundancy into infrastructure networks is highlighted by the performance of the gas network in Christchurch. After both the September 2010 and February 2011 earthquakes, one plant was forced shut down, but service levels were able to be retained due to capacity in the available remaining plants. Conversely, Christchurch is heavily reliant the waste water treatment plant at Bromley and so problems here had a direct and severe impact on the city’s ability to cope with waste water. Internal redundancy within the water system was very instrumental in L’Aquila for speeding up and optimising the restoration of the service. The restoration of the water provision was in fact gradually operated starting from the less affected zones and/or the zones with a strong need for reactivation (e.g. critical buildings). The partial restoration of the water distribution was possible thanks to the presence of secondary networks and of a shutter system that allowed the exclusion of areas where the water supply was not urgently needed.

5.2 Organisational performance

All the infrastructure management organisations in Christchurch were able to maintain their operation after the earthquakes demonstrating organisational robustness. However in the case of Orion, the electricity distribution company, it was necessary to vacate their main operations centre due to damage after the February earthquake, but they were able to move to an alternative facility to minimise disruption demonstrating redundancy in their operational facilities.

When frequencies in the South Island (NZ) electric power network increased after the February 2011 earthquake, Transpower were able to implement a plan to balance the load by redistributing energy to North Island (Kwasinski et al. 2014). The ability to make real-time decisions on infrastructure network is critical to their protection in the immediate aftermath of an event and also to ensure timely and safe repair and recovery operations. Access to GIS and to accurate geospatial network data contribute significantly to support rapid emergency decision-making, as was evident in the recovery of the gas network in L’Aquila.

In Christchurch there was significant damage to major distribution lines in the east of the city, causing significant outages to residents in that area. However Orion were able to act quickly to restore as much of this service as possible by using available equipment in stores to construct a new temporary overhead line. The work which would normally take 6 months was completed within 1 week (Giovinazzi et al. 2011). Other examples of rapid-response actions to restore access to services were implemented by Christchurch City Council, which was able to set up temporary desalination plants to provide fresh water and also provided temporary sewage facilities including chemical and portable toilets.

In L’Aquila, factors that contributed to the fair resilience observed, included the prompt availability of both physical and human resources and a thoughtful prioritisation of the recovery operations. In fact the reinstatement of the gas service delivery, following the reactivation operations after the gas network shut-off and the required repair/replacements activities, were closely monitored by the network operator, in collaboration with the Lifeline Function of the L’Aquila Emergency Operation Centre (operated by the Italian Civil Protection).

5.3 Social performance

In both Christchurch and L’Aquila, emergency management decisions were made to shut down the gas systems in order to prevent the risk of fire or explosion. Even though parts of these systems were able to withstand the earthquake, uncertainty regarding the earthquake impact meant that although it is the goal of infrastructure managers to restore services as quickly as possible, safety must still come first. This was especially important in the Christchurch CBD where damage to the water system hindered the firefighting services in their capacity to fulfil their role.

In both Christchurch and L’Aquila, the gas systems have not yet recovered full functionality and their numbers of customer connections are down from pre-earthquake levels but this does not necessarily mean that they are not resilient. A major reason for the drop in customer connections is that the commercial centres of both cities are still at very early stages of recovery with a fraction of the number of businesses operating. Although they have not been able to regain all their customers, both Enel (in L’Aquila) and Contact Energy (in Christchurch) have achieved as much as they can in terms of recovery whilst waiting for reconstruction to move forward. In other words, it is the lack of rapidity in urban reconstruction rather than in infrastructure resilience that is driving the ongoing reduced functionality. Therefore consideration of the rapidity criteria for infrastructure resilience should take into account the post-earthquake state of the communities they serve and not just the pre-earthquake level of functionality.

5.4 Interdependencies

In both the September 2010 and February 2011 earthquakes, failure of gas plants occurred because of loss of supply from other utilities (electric power and water respectively), signifying the impacts of operational interdependencies between systems. Pumping stations in the water system also experienced loss of power supply and whilst some were able to make use of back-up generators this was not the case for all of them. It was fortunate in these cases that redundancy amongst facilities prevented any functional impact arising. The case of the gas plant in the September 2010 earthquake was particularly interesting since interdependency had been considered with the presence of the back-up generator but unfortunately this also failed. This example shows that whilst it is essential to consider interdependencies, the provision of back-up supplies does not replace the need to build redundancy into the system. The two considerations should be complementary rather than alternative options. Physical interdependencies also caused problems as demonstrated by the damage to electrical cables due to bridge failure at Dallington in Christchurch. Whilst there is no operational or physical relationship between the gas network and road network in Christchurch, damage to roads after the earthquakes created difficulties for emergency inspection and repair teams. Similar problems were observed due to failure of radio communications. These are examples of logistical interdependencies that do not routinely exist, but only materialise in emergency situations, and show that interdependencies need to be considered dynamically.

As for the L’Aquila earthquake, operational interdependencies were observed between the electric power and the waste water systems, i.e. the waste water treatment plant lost supply from the electric power network for 3 days (Tang and Cooper 2009).

Logistical interdependencies were as well observed, i.e. the reduced functionality of the urban road network (mainly caused by the presence of debris from damaged/collapsed buildings and/or precarious buildings adjacent to the roads) slowed down the emergency response and reinstatement activities for the other networks. Similarly construction sites that needed to put in place on the roads to address urgent repairs operations required to the gas and water pipes contributed to reduce the capacity of the already overloaded road network.

6 Conclusions

The post-earthquake assessment of critical infrastructure systems is currently hindered by the lack of a consistent approach to reporting observations, which makes it difficult to compare performances across earthquakes and also between observations and model predictions. This paper has taken the first steps towards the definition of a standardised survey tool by proposing suitable taxonomy and physical damage descriptors, based on the existing literature for assessing risk to infrastructure. The taxonomy and damage descriptors have been applied to the L’Aquila and Canterbury earthquakes, collating information from a variety of sources.

The observed damage data has shown that strong motions in both locations, and liquefaction in Canterbury, led to significant damage and disruption to infrastructure systems. Despite this, studies on the electric power (Kongar et al. 2014) and road (Bocchini et al. 2014) networks in Christchurch and on the gas network in L’Aquila (Esposito et al. 2013) have shown that existing predictive tools such as HAZUS have over-predicted damage from these events and therefore may need to recalibrated and tuned for modelling future scenarios, depending on the nature of the risk assessment being conducted. It is to inform this type of analysis that rigorous and standardised data collection of observed infrastructure damage is important, since in many cases, as shown in this paper, the data is not available to the right quality. The differences in the reporting quality of damages between L’Aquila and Christchurch make a direct comparison between the two events infeasible, which confirm the value that a standardised reporting tool would give.

However, infrastructure systems do not exist as purely physical assets. They are systems that provide essential services to communities and so their performance has a direct impact on the resilience of that community to an earthquake and the ability of people to function and live comfortably. Therefore the importance of post-earthquake infrastructure assessments is not just for physical damage, but also systemic performance and other aspects of infrastructure resilience as described by Bruneau et al. (2003). This paper has shown examples of good technical performance due to robustness of assets (e.g. PE pipelines, anchored substations in Christchurch) but this has often been the result of resilient design and learning from the lessons of previous earthquakes and there have been other success stories at organisational and societal level. It is important that these, as well as the challenges and disappointments are documented so lessons can be passed on to managers in other at-risk regions.

The two events have provided a wealth of information for infrastructure operators, emergency managers and engineers to discuss and learn from, but this would have had even greater value if the information was presented through a consistent framework. This paper has presented some initial proposals on how to achieve that in relation to taxonomy and damage description. However, there needs to be a more global agreement to achieve the objective of a consistent framework and more consideration must be given to tools for documenting systemic functional impacts as well as organisational performance and societal consequences.