otechnique 60, No. 7, 505543 [doi: 10.1680/geot.2010.60.7.505] ORourke, T. D. (2010).

Ge

Geohazards and large, geographically distributed systems

T. D. O RO U R K E A general classication for scale in geotechnical engineering is used to explore the modelling of large, geographically distributed systems and their response to geohazards. Both component and network performance are reviewed. With respect to components, prototypescale experiments of underground pipeline response to abrupt ground deformation are described, including control of soil properties, soilpipeline interaction, and performance of high-density polyethylene pipelines. Direct shear (DS) apparatus size is shown to have a signicant effect on DS strength, and the most reliable DS device is identied from comparative tests with different equipment. MohrCoulomb strength parameters for partially saturated sand are developed from DS test data and applied in nite element simulations of soilpipeline interaction that show excellent agreement with prototypescale experimental results. Apparent cohesion measured during shear failure of partially saturated sand is caused by suction-induced dilatancy. With respect to networks, the modelling of liquefaction effects on the San Francisco water supply is described, and a case history of its successful application during the Loma Prieta earthquake is presented. The systematic analysis of pipeline repair records after the Northridge earthquake is used to identify zones of potential ground failure, and correlate pipeline damage rates with strong ground motion. Hydraulic network analyses are described for the seismic performance of the Los Angeles water supply, with practical applications for emergency response. The effects of Hurricane Katrina are reviewed with respect to the New Orleans hurricane protection system, Gulf of Mexico oil and gas production, and interaction between electric power and liquid fuel delivery systems. The sustainability of the Mississippi delta is discussed with regard to ood control, maintenance of wetlands and barrier islands, and catastrophic change in the course of the Mississippi River. ne rale pour le chelle en On utilise une classication ge nierie ge otechnique an dexplorer la mode lisation de inge ` mes a ` distribution ge ographique, et leur grands syste action en pre sence de risques ge ologiques. On se re penche sur les performances des composants et du re riences seau. En ce qui concerne les composants, des expe action de conduites dhydrocarbures sur prototype de re ` des de formations soudaines du sol sont souterraines a crites, y compris le contro le de proprie te s du sol, les de interactions sol - conduites dhydrocarbures, et les perfor thyle ` ne a ` haute densite . On mances de conduites en polye montre que les dimensions des appareils a ` cisaillement de sistance au cisailledirect ont un effet signicatif sur la re ment direct, et on identie le dispositif de cisaillement s avec direct le plus able au moyen dessais compare rents e quipements. On de veloppe des parame ` tres de diffe sistance de Mohr-Coulomb pour du sable partiellement re , sur la base de donne es dessai a ` cisaillement sature ` des simulations aux e le ments direct, que lon applique a nis dinteractions sol conduite dhydrocarbures pre sultats sentant une excellente correspondance avec des re rimentaux sur prototypes. La cohe sion apparente, expe e au cours dune rupture par cisaillement du sable mesure , est cause e par une dilatance induite partiellement sature seaux, on de crit par laspiration. En ce qui concerne les re lisation des effets de la lique faction sur les fourla mode sente une e tude nitures deau de San Francisco, et on pre ussie au cours du tremblement de cas dune application re de terre de Loma Prieta. On fait usage de lanalyse matique des dossiers de re paration des conduites syste ` la suite du tremblement de terre de dhydrocarbures a Northridge pour identier des zones potentielles dacci terminer lendommagement potendents de terrain, et de sence de tiels de conduites dhydrocarbures en pre s du terrain. On de crit des analyses mouvement prononce seau hydraulique pour les performances sismiques du re seau de canalisations de fourniture deau de Los du re Angeles, avec des applications pratiques pour des interventions en cas durgence. On examine les effets de ` me de protection louragan Katrina en fonction du syste ans contre les ouragans, des installade la Nouvelle Orle trole et de gaz dans le Golfe du tions de production de pe seaux de fourMexique, et de linteraction entre les re lectricite et les syste ` mes de fourniture de nitures de du delta combustibles liquides. On discute de la viabilite du Mississipi en ce qui concerne la lutte contre les inondations, lentretien des zones humides et des les` re, et la mutation catastrophique du cours du barrie Mississipi.

KEYWORDS: buried structures; case history; earthquakes; full-scale tests; laboratory equipment; laboratory tests; liquefaction; numerical modelling; partial saturation; sands; seismicity; shear strength; soil/structure interaction; suction

Manuscript received 2 February 2010; revised manuscript accepted 15 April 2010. Discussion on this paper closes on 1 December 2010, for further details see p. ii. School of Civil and Environmental Engineering, Cornell University, USA.

INTRODUCTION Geotechnical engineers play a critical role in managing the performance of large, geographically distributed systems that are affected by geohazards such as earthquakes, oods, hurricanes and landslides. Systems, such as water supplies, levees and gas and liquid fuel supply networks, may cover thousands of square kilometres and be subject to many 505

506

OROURKE distances within normal line of sight. At the mega scale, geotechnical engineering is applied, for example, to river, coastal and offshore systems, where ood protection structures, transportation networks, and energy production and delivery systems are constructed and managed. There are many scales explicitly or implicitly embodied in models that represent large, geographically distributed systems. For example, pipeline system response to an earthquake is addressed at a minimum of two levels (ORourke et al., 2008), involving: (a) component performance, for which soilstructure interaction under earthquake loading is evaluated; and (b) system performance, for which the integrated behaviour of the network is assessed. Such multi-scale models involve trade-offs between the detail required for accuracy and the simplication needed for computational efciency and practical applications. Models of large, geographically distributed systems generally focus on kilo- and mega-scale operations that properly account for network ow laws and the spatial variability of both the system infrastructure and geohazard effects. At the component level, relatively simple models may be based on expert opinion or empirical correlations between capacity and demand. More complex component models are based on nite element simulations or fragility curves that express the probability of failure as a function of loading intensity (e.g. Shinozuka et al., 2000; Kafali & Grigoriu, 2007). The most important criteria for modelling large, geographically distributed systems are model validation with respect to actual eld performance, and critical review and acceptance by system operators. These validation and acceptance criteria establish the level of practical worthiness necessary for credible results. This paper explores nine orders of magnitude in the modelling of large, geographically distributed systems, from the behaviour of partially saturated sand to the performance of the water supply system in Los Angeles and the ood protection system in the Mississippi River delta. The concept of scale is used as a unifying theme to show how the shear strength of sand at the milli-scale level is linked through soilstructure interaction of pipeline components at the macro-scale level to evaluate the earthquake response of a regional water supply system at the mega-scale level. The effects of Hurricane Katrina are used to show how geotechnical problems are shaped by uvial, deltaic and cyclonic processes that are distributed over many thousands of kilometres, with enormous impact on large, geographically distributed systems. GEOHAZARDS Table 2 provides a summary of major geohazards, with information about their spatial characteristics as well as their

different ground response and geotechnical failure mechanisms. The geotechnical factors affecting system behaviour have broad implications for life safety and regional economic stability. The paper begins with a discussion of geotechnical scales, and how scale affects both problem denition and solutions. It then examines how soilstructure interaction of components is modelled for extreme loading conditions, by describing laboratory tests of underground pipeline response to ground rupture. The tests are used to illustrate how experiments performed at prototype scale under controlled conditions in the laboratory improve our understanding of soil structure interaction, and lead to improvements in modelling of soil behaviour. The geotechnical factors affecting regional system response to geohazards are examined with reference to earthquake effects on the San Francisco and Los Angeles water distribution networks, as well as hurricane effects on the New Orleans hurricane protection system. The paper describes the modelling procedures for simulating water supply performance during earthquakes, and explores how modelling results are used to address liquefaction and ground failure hazards, develop relationships between earthquake motion and underground pipeline damage, and make riskbased decisions about the emergency operation of water supplies. The effects of Hurricane Katrina on New Orleans and Gulf of Mexico oil and gas infrastructure are described, and the implications of mega-scale problems associated with hurricane and river ooding, coastal restoration and catastrophic change in river location are reviewed with respect to critical infrastructure performance. SCALE IN GEOTECHNICAL ENGINEERING The physical dimensions involved in geotechnical engineering span 15 orders of magnitude when compared on a metric scale, as shown in Table 1. The table is organised in accordance with a similar scaling procedure proposed by Chong & Davis (1999) for conceptualising research in civil and mechanical engineering. It provides a generalised model in which there are six divisions of scale, ranging from nano scale (109 m) at the molecular level to mega scale (106 m) at the level of geographic regions. For each division of scale, examples of natural features and engineering applications are given. For example, clay surfaces and double-layer effects at the nano scale inuence the strength, volume change and hydraulic characteristics of soils (e.g. Mitchell & Soga, 2005), and examples of engineering applications that exploit nano-scale characteristics include electro-osmosis as well as chemical and bio-remediation. Most geotechnical testing and characterisation occur at the milli and macro scales, which conform essentially to human dimensions and

Table 1. Summary of scales in geotechnical engineering Nano 109 m Molecular Clay surfaces Double-layer effects Micro 106 m Micrometres Clay Silt Milli 103 m Millimetres Sand Gravel Cobbles Suspension grouts Reinforcing strips Polymer grids Macro 100 m Metres Slopes Soil deposits Foundations Retaining walls Pipe/conduits Kilo 103 m Kilometres Aquifers Reservoirs River crossings Bridges Dams Tunnels Levees Mega 106 m Regions River, coastal and offshore systems

Cation exchange Electro-osmosis Chemical and bio-remediation

Water supplies Transportation networks Electric power and fuel systems

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Table 2. Geohazard spatial characteristics and effects Geohazard Earthquake Tsunami Spatial characteristics Area of 102 to 105 km2 with signicant shaking and ground failures Travel distances of 1 to 104 km, affecting 10 to 102 km of coastline and waterfront Eye of storm covers 10 to 102 km2 . Area of high winds and storm surge may cover 102 to 106 km2 10 to 103 km along main river courses and from 10 to 106 km2 within oodplains Area affected by volcanic activity varies from 10 to 105 km2 Large regional effects of 105 to .106 km2 Local to regional settlement affecting 10 m2 to 103 km2 Direct effects Strong ground shaking Faulting, landslides, liquefaction, and consolidation of loose soils Flooding Impact from ood debris Erosion and removal of soil, vegetation, dwellings, and infrastructure Flooding and high winds Impact from airborne and waterborne debris Erosion and destruction of wetlands, soils, and waterfront facilities Inundation Flood loads and impact of ood debris Soil erosion, scour, and undermining of structures Lava and pyroclastic ows Volcanic ash and projectiles Lahars and oods Noxious gases Loss of water and vegetation Loss of crops Increased re hazard Soil erosion and shrinkage Settlement, change of grade, and lateral ground movement Flooding Reduced aquifer storage Growth faults with abrupt vertical and horizontal ground displacement Lateral ground movement, settlement, and change in grade. Movements may be rapid or develop gradually. Landslide dams Indirect effects

507

Threat to all civil infrastructure, including buildings, waterfront facilities, transportation and other lifeline systems Threat especially to coastal and waterfront facilities. Large run-up and inundation of coastal buildings and infrastructure

Hurricane and typhoon

Threat especially to coastal and waterfront facilities. Wind and ood damage to buildings, ood protection systems, and lifelines. Transmission towers and electrical substations vulnerable Threat especially to river crossings and oodplain infrastructure, such as bridges, levees, port facilities, dwellings, and lifeline systems Infrastructure damage from high heat, re, ooding, debris ows, projectile impact and ash falls. Saturated ash loading and obstruction of infrastructure Loss of reservoirs, watersheds, food supply and industry. Wildres and urban res. Slope failure due to vegetation loss Adverse effects on stormwater drainage, canals and transportation systems. Differential settlement and lateral deformation of buildings and lifelines. Flooding due to subsidence is a threat to all civil infrastructure

Flood

Volcano

Drought

Subsidence

Landslide

Volume varies from , 10 m3 to . 107 m3 Largest historic volumes of 23 km3

Catastrophic damage to buildings and infrastructure when movements are large and rapid. Landslide dams threaten downstream infrastructure

direct and indirect effects. Geohazards have attracted substantial public and policy-making attention, with nearly 90 000 dead and missing, and substantial infrastructure destruction in China from the 2008 Sichuan earthquake (EERI, 2008; Stone, 2009); over 228 000 people killed by the 2004 Indian Ocean tsunami (Iwan, 2006; Cosgrave, 2007); and over $100 billion in direct losses (Jordan & Paulius, 2006) and nearly 2000 dead and missing (IPET, 2008a) from Hurricane Katrina in 2005. At the time of submitting this paper, the dead and missing after the 2010 Haiti earthquake are estimated at approximately 200 000, with nearly complete destruction of the physical and governmental infrastructure of its capital city, Port-au-Prince. Geohazards have a strong inuence on professional practice and research in geotechnical and civil engineering. Moreover, they increase the scope and extent of geotechnical interests from sitespecic practices to regional concerns and interdisciplinary policies. Geohazards have generated substantial interest in lifeline systems. Lifelines include electric power, gas and liquid fuels, telecommunications, transportation, waste disposal, and water supply (ORourke et al., 2001). This grouping can easily be extended to include ood protection and levee systems for coastal and river communities. Taken individually, or in the aggregate, such systems are intimately linked with the economic well-being, security and social fabric of the communities they serve. Thinking about lifelines helps

clarify the resources and services that are essential for community resilience, and encourages an exploration of the interdependences among critical infrastructure systems (ORourke, 2007). The geohazards in Table 2 include earthquake, tsunami, hurricane and typhoon, ood, volcano, drought, subsidence and landslide. All represent a life safety threat, but only select effects on the built and natural environments are listed in the table. Most geohazards affect broad areas, primarily in the range of 102 to 105 km2 . Coverage may be restricted to zones as small as 10 m2 owing to local subsidence and slope failure, or may involve a region exceeding 106 km2 in the case of drought. Special attention is given to earthquakes and hurricanes in this paper. The widespread effects of earthquakes are especially well documented. The 1906 San Francisco earthquake, for example, involved signicant ground shaking throughout an area of about 48 000 km2 (Lawson et al., 1908; Ellsworth, 1990). Seismic ground waves generated by the 181112 New Madrid earthquakes caused damage in an area exceeding 500 000 km2 (Johnston and Schweig, 1996). Earthquake effects on transportation, liquid fuel, electric power and water supply systems are covered extensively in the technical literature (e.g. ORourke et al., 2004). The economic impact of earthquakes is well illustrated by the 1995 Kobe earthquake, with direct losses of $100150 billion (Hamada et al., 1995).

508

OROURKE structure is removed from direct observation unless uncovered, and its state of repair and its proximity to other structures are often unknown. Third, congestion increases risk due to proximity. Damage to one facility, such as a cast iron water main, can cascade rapidly into damage in surrounding facilities, such as electric and telecommunication cables and gas mains, with system-wide consequences. Soil surrounding critical underground infrastructure is frequently both the perpetrator and the mediator of loading that can affect the systemic performance of an entire city. LARGE-SCALE EXPERIMENTS A signicant trend in geotechnical engineering has been the implementation of large-scale testing facilities for soil structure interaction, such as those at the Japanese National Research Institute for Earth Science and Disaster Prevention that have been used to characterise soilpile interaction during liquefaction (Tokimatsu & Suzuki, 2004) and the large-scale split box experiments at the George E. Brown, Jr Network for Earthquake Engineering Simulation (NEES) equipment site at Cornell University (e.g. Palmer et al., 2006; ORourke & Bonneau, 2007). The large-scale facilities allow for testing at the macro-scale level (see Table 1), so that conditions in the eld can be simulated reliably under laboratory control. It is not possible to model with accuracy the soil displacement patterns at all potentially vulnerable locations. It is possible, however, to set an upper bound on deformation by simplifying spatially distributed ground deformation as abrupt soil movement. Detailed studies of fault deformation disclose that abrupt soil rupture and offsets are indeed recurrent patterns at active faults (Bray et al., 1994). They thus establish a baseline to evaluate soilpipeline interaction representative of upper-bound ground movements for earthquakes, landslides, subsidence and ood undermining. As shown in Fig. 2, split-box testing has the capability of imposing abrupt soil displacements on buried pipelines consistent with those at fault crossings and the margins of lateral spreads and landslides. Relative displacement is generated along a movable interface between two test basins, or boxes, containing soil and the buried pipeline. The pipeline is buried in soil that is placed, compacted and tested according to eld construction practice. The dimensions of the experimental boxes are selected on the basis of computational modelling and previous test experience to minimise the inuence of the test facility boundary effects. Figure 3 shows the current generation of split-box test basin that utilises between 90 and 100 metric tonnes of partially saturated sand per test and 1.2 m of strikeslip displacement. In the gure, the displacement was provided by two hydraulic structural actuators with load capacities of 445 kN tension/650 kN compression and a one-way stroke of 1.28 m. Other tests have been performed with four hydraulic actuators positioned to pull the movable section of the test basin with a combined capacity of 1.5 MN in tension. As explained by ORourke et al. (2008), a special conveyor system was fabricated to move large quantities of soil and facilitate soil placement in about 4 to 5 days. Soil for a typical experiment was placed in seven 200 mm lifts and one top lift of 100 mm. Each lift was compacted with two passes of a gasoline-powered plate tamper. The soil was a glacio-uvial sand, referred to as RMS graded sand, produced in accordance with New York State specications for concrete sand, and representative of sand used for backll and engineered construction. Between 240 and 320 measurements of dry unit weight and moisture content were recorded for each set-up. Dry unit weight, dry , was measured in situ using a nuclear density gauge according to ASTM D6938-

A hurricane or typhoon is a tropical cyclone with maximum sustained wind speed of at least 119 km/h (e.g. Emanuel, 2005; Rauber et al., 2005). Its cloud shield diameter is typically on the order of 800 km, although diameters of the largest storms can exceed twice this value (Fitzpatrick, 2006). The worst effects are associated with storm surge ooding, which historically accounts for 90% of cyclone-related deaths (Longshore, 1998). In 2005 Hurricanes Katrina and Rita were responsible for disrupting 100% and 94% of all Gulf of Mexico offshore oil and gas production respectively (MMS, 2006). Hurricane Katrina caused enormous damage in the New Orleans area, resulting in the greatest natural disaster loss in US history, as well as controversy about the performance and failure mechanisms of the regional hurricane protection system (e.g. ASCE Hurricane Katrina External Review Panel, 2007; Christian, 2007; Seed et al., 2008a).

EXTREME LOADING CONDITIONS Extreme loading associated with geohazards often occurs with large plastic, irrecoverable soil deformation. Such deformation may involve geometric changes in the soil mass, such as localised shear rupture, heave, and void formation. Soilstructure interaction under these conditions is frequently characterised by a peak, or maximum, interaction force. The extreme loading of soils occurs collaterally with the extreme loading of structures. Examples include soil structure interaction associated with pipelines subjected to fault rupture, piles affected by landslides, and underground facilities exposed to lateral soil movement, ooding and undermining. Structural response often entails large plastic, irrecoverable deformation with material and geometric nonlinearities. Hence analytical and experimental modelling for soilstructure interaction under extreme conditions requires the coupled post-yield simulation of both soil and structure. Such behaviour imposes signicant demands on modelling, thus requiring large-scale experimental validation of analytical models to improve the simulation process. Extreme loading, especially in conjunction with geohazards, often affects large lifeline systems. Consider, for example, Fig. 1, which is a photograph of the corner of Wall and Williams Streets in New York City during the 1920s. The photo illustrates at least three important features of the urban environment. First, much of the critical infrastructure is located underground, and its fate is intimately related to that of the surrounding ground. Second, underground infra-

Fig. 1. Underground infrastructure at Wall and Williams Streets in New York City, 1920s (photograph provided by the Consolidated Edison Company of New York, Inc., New York, NY)

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Special trench and backfill at fault crossing Compacted sand Welded steel pipeline Pipe trench cross-section (a) Straight pipe

509

Fixed box

Compacted sand Welded steel pipeline

Elbow

Buried pipeline Permanent ground deformation (PGD) (b)

Elbow

Fixed box Pipe with elbow (c)

Fig. 2. Simulation of ground rupture effects on pipelines by split-box tests: (a) PGD effect on buried pipelines; (b) PGD effect on buried pipelines with elbows; (c) experimental concepts

32 m 12 m 66 m

23 m Actuators

Fig. 3. Large-scale split-box test basin at the Cornell University NEES equipment site

07b (ASTM, 2008a), and moisture content was measured in situ according to ASTM D2216-98 (ASTM, 2008b). Typically, from ve to seven dry unit weight determinations were made per cubic metre of soil. Figure 4 presents the grain size distribution curve for the RMS graded sand, which is angular to sub-angular with non-plastic nes. Examination assisted by microscope showed approximately 71% by volume of siltstone, ne100 Angular to sub-angular 29% quartz 71% sedimentary rock fragments 50

0 10

01

001

Particle diameter: mm

Fig. 4. Grain size distribution curve for RMS graded sand used in the large-scale tests

grained sandstone, shale and limestone fragments, and 29% quartz grains. The mean grain size D50 for the soil is 0.67 mm, and the coefcient of uniformity Cu is 2.83. The peak friction angle ds-p associated with dry measured by the nuclear density gauge was determined from regressions that had been developed between dry and ds-p using direct shear (DS) tests for partially saturated sand (ORourke et al., 2008). The value of ds-p is reported in terms of total stress, because suction in the moist sand (and thus effective stress conditions) is not measured directly. For 9s-p . dry sand, ds-p d Figure 5 shows plots presented by ORourke et al. (2008) of the mean dry unit weight dry , gravimetric water content w and peak friction angle ds-p of the sand for each lift with respect to depth for ve full-scale split-box pipeline experiments. As shown in the gure, a remarkable degree of control was achieved over the experimental soil properties for many tests with large volumes of material. The preponderance of the measurement data show placement of the sand in the range of dry 15.515.8 kN/m3 , w 3.54.5% and ds-p 39408. Statistical control procedures described by Trautmann et al. (1985) were applied to determine condence intervals for the sand properties based on mean characteristics of the entire soil mass. For example, the 95% condence intervals on the mean dry and ds-p for any given test were 0.020.05 kN/m3 and 0.100.198 respectively. Fig. 4 illustrates the spread in soil properties lift by lift among several tests, whereas the condence intervals pertain to overall average values for each individual test. Partially saturated sand is different from saturated or dry sand in that it is affected by matric suction, um . The presence of um promotes apparent cohesion in sand, and increases its dilatancy relative to dry and saturated sand at the same dry . To help evaluate the properties of the partially saturated test sand, soil water retention curves (SWRCs) were obtained with Tempe cells, and tensiometer measurements were taken during large-scale tests to generate relationships between w and um . The Tempe cell, described in ASTM D6836-02 (ASTM, 2008c), was used to develop SWRCs by desorption (drying), while tensiometers, described by Dane & Topp (2002), were used to measure um in situ during large-scale tests.

% Passing

510
Test 1 20 155158 kN/m3 16

OROURKE
Test 2 Test 3 Test 4 Test 5

3545%

390400

Depth: m

12

08

04

0 152

156 160 d: kN/m3

164 002

003

004
w

005

006

007 380

390 400 410 ds-p: degrees

420

Fig. 5. Mean dry unit weight, water content and peak friction angle by lift against depth for ve largescale tests (ORourke et al., 2008)

Figure 6 shows a plot of w against um , comparing Tempe cell and tensiometer measurements for partially saturated RMS graded sand at dry 15.7 kN/m3 with equations proposed by Pham et al. (2005) for the main drying curve (MDC) and main wetting curve (MWC). Pham et al. developed equations to represent the MDC and MWC for sand based on a known SWRC and key soil properties. The relationships presented by Pham et al. are for volumetric water content, but are adapted here for w. The Tempe cell desorption and tensiometer data plot closely with respect to the MDC and MWC respectively. The tensiometer measures a vacuum pressure, generated as water is transported across the tensiometer porous stone into the surrounding soil. This mechanism of measurement is analogous to generating an SWRC by adsorption (wetting), and thus the tensiometer data follow the MWC. The RMS graded sand naturally drains in the large-scale test basin to w % 4%, which is about equal to the residual water content, as shown in the gure. SHEAR STRENGTH TESTING OF SAND Over 700 DS tests were performed to quantify the DS behaviour of sand for the soilpipeline interaction tests. Such testing corresponds to the milli-scale level (see Table 1), and was intended to develop DS testing procedures for the accurate determination of strength. Of primary interest were the effects of DS box dimensions. It has been observed
024 MDC (Pham et al., 2005) MWC Tempe cell Tensiometer Main wetting curve, MWC Main drying curve, MDC Residual water content 004 d 157 kN/m3 0 001 01 1 Matric suction, um: kPa 10 100

020 016 012 008

Fig. 6. Soil water retention curve for RMS graded sand used in the large-scale tests

(e.g. Dietz, 2000; Cerato & Lutenegger, 2006) that small ratios of box dimension to particle ratios lead to boundary constraints on DS behaviour and overestimation of strength and dilation parameters, while large ratios lead to progressive failure and underestimation of the same parameters. Moreover, to minimise errors, several investigators have modied the DS equipment to restrain upper frame and load pad rotation (e.g. Jewell & Wroth, 1987; Shibuya et al., 1997; Lings & Dietz, 2004). Tests were performed with DS testing devices, having nominal interior horizontal dimensions of (a) 60 mm 3 60 mm, (b) 100 mm 3 100 mm and (c) 300 mm 3 300 mm, for which the key horizontal dimension L is 60 mm, 100 mm, and 300 mm respectively. Photographs of the 60 mm, 100 mm and 300 mm test boxes are shown in Fig. 7. The test programme is described in detail by Olson (2009), and only the salient features are summarised here. The 60 mm DS tests were performed using a conventional Wykeham Farrance DS testing apparatus, which conforms to the specications of ASTM D 3080-04 (ASTM, 2008d) and BSI 1377-7 (BSI, 1990). The device had a conventional brass lower box and an upper box machined from maple wood. The wooden upper frame weighed 0.9 N in comparison with the 13.0 kN brass upper frame, thus minimising the vertical load transferred from the device to the soil failure plane. The 100 mm and 300 mm DS tests were performed using a custom-fabricated DS testing apparatus, designed in accordance with the characteristics of the 100 mm test device described by Dietz (2000) and Lings & Dietz (2004). Load is applied in this device to a pair of yokes at the midpoint of the upper frame sidewalls, such that the application of shear force is coincident with the central horizontal plane of the soil specimen. Great care was taken to prepare the soil with proper dry , as described by Olson (2009). Strict soil placement and tamping procedures were employed to promote uniform conditions with respect to soil fabric and in situ stress. Numerous tests were performed to evaluate the inuence of gap size (separation between upper and lower DS test frames) and edging (rubber skirts to prevent extrusion of soil through the gap). A 1 mm gap without edging was adopted for the tests. It was found to be sufciently small that soil loosening and deterioration did not occur in the separation between the upper and lower boxes, yet sufciently large that measurements of peak shear strength and dilation were

Gravimetric water content, w

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

511

L 60 mm

L 100 mm (b)

(a)

L 300 mm

(c)

Fig. 7. Photographs of DS test devices: (a) 60 mm test box; (b) 100 mm test box; (c) 300 mm test box

not amplied by constraints imposed by the testing apparatus. For dry sand tests the peak friction angle d 9s-p was obtained from the peak shear stress p 9 for a given normal stress N 9 as tan d 9s-p p 9 N 9 (1)

The peak angle of dilatancy, p , was obtained from the maximum change of vertical soil displacement, v y , relative to the horizontal displacement, v x , between the upper and lower frames of the DS test apparatus, using tan p dv y dv x (2)

Figure 8 shows d 9s-p against dry for RMS graded sand tested at N 9 22 kPa. Tests with sand having different grain size characteristics were performed (Olson, 2009), but only

the test results for RMS graded sand are summarised herein. Sand placed at low dry < 15.6 kN/m3 was contractive throughout the entire test. The data trend for sand at low dry differs markedly from the linear trend at higher unit weights, and they are not included in the linear regression ts to the data. In general, the regression line for the 60 mm box is 468 higher than that for the 100 mm box, and 678 higher than that for the 300 mm box. The 100 mm box trend line has a coefcient of determination r 2 of 0.98, higher than the values of 0.90 and 0.89 for the 60 mm and 300 mm boxes respectively, providing an indication of the improved consistency of data obtained with the 100 mm box. Fig. 9 shows p against dry for RMS graded sand. The 60 mm box line plots 378 higher than those for the 100 mm and 300 mm boxes, which are statistically indistinguishable from each other. The trend lines all have relatively high r 2 values, with the 100 mm box showing the highest r 2 0.94.

512
55 Linear fit: L 60 mm (dry 156 kN/m3) y 557x 5159 n 26, r 2 090 Linear fit: L 100 mm (dry 156 kN/m3) y 659x 7321 n 11, r 2 098

OROURKE device was discontinued. Tests were performed with the 100 mm device, fabricated in lightweight aluminium according to the design of Lings & Dietz (2004), and shear strength values obtained with this device are presented in the remainder of this paper. SHEAR STRENGTH OF SAND The DS test does not provide a peak stress state at maximum obliquity with respect to the Mohr circle: that is, it does not represent a point on the Mohr circle of stress that is tangent to the MohrCoulomb failure surface. Constitutive laws that utilise the MohrCoulomb failure surface are readily available in software, such as ABAQUS and FLAC. They require specication of the maximum obliquity effective stress friction angle p 9s (also referred to as the plane-strain friction angle) for plane-strain problems of soilstructure interaction. At peak DS strength for dry granular media, researchers have shown coaxiality of stress and incremental strain, meaning that the principal axes of stress and incremental strain coincide (e.g. Lings & Dietz, 2004; Bolton, 1986). Figs 10(a) and 10(b) are Mohrs circles of incremental strain and stress respectively. Assuming coaxiality, and recognising that the horizontal axis is a direction of zero linear incremental strain (zero extension), one can use the dilation angle and Mohrs circle of stress to develop a relationship between 9 , and the plane-strain values of the DS values of p 9 and N 9N . Davis (1968) rst derived shear, p 9s , and normal stress, p the equation linking the parameters shown in Fig. 10(b) as tan d 9s cos sin p 9s 1 sin sin p 9s (3)

50

45

ds-p: degrees

40

35

Linear fit: L 300 mm (all dry) y 587x 6239 n 9, r 2 089 L 60 mm L 100 mm L 300 mm Linear fit: L 60 mm Linear fit: L 100 mm Linear fit: L 300 mm 16 17 dry: kN/m3 18 19

30

25

20 15

Fig. 8. d 9 s-p against dry for three different DS test device sizes with RMS graded sand
30

25

L 60 mm L 100 mm L 300 mm Linear fit: L 60 mm Linear fit: L 100 mm Linear fit: L 300 mm Linear fit: L 60 mm (dry 156 kN/m3) y 863x 13275 n 26 r 2 091 Linear fit: L 100 mm (dry 156 kN/m3) y 698x 10948 n 11, r 2 094 Linear fit: L 300 mm (All dry) y 730x 11461 n 9, r 2 091 16 17 dry: kN/m3 18 19

which at the critical state ( 0) becomes tan d 9s sin c 9rit (4) in which c 9rit is the critical-state friction angle. Bolton (1986) and Jewell & Wroth (1987) provide similar derivations. It is assumed that the coaxiality of incremental strain and stress applies for partially saturated RMS graded sand. Given the low values of suction measured for the test sand, it is likely that the relative orientations of stress and incremental strain are not materially different from those for dry sand. Moreover, it is shown later that strength and volume change characteristics evaluated on the premise of coaxiality support a rational mechanics-based explanation of observed behaviour and strength parameters that leads to a very favourable comparison between analytical and experimental results. Figure 11(a) shows against for a partially saturated sand with c and . From this gure it can be seen that the centre and top of the Mohr circle of stress is related to (p /N ) through p , such that 1 3 (5) N p tan p 2 p 1 3 (6) 2 cos p in which 1 and 3 are the major and minor principal stresses respectively. If one develops a linear regression for (1 + 3 )/2 (centre of the Mohr circle) and (1 3 )/2 (radius of the Mohr circle) from the DS data, one obtains a plot with intercept c and slope tan , as shown in Fig. 11(b). The plane-strain failure envelope for maximum obliquity is also illustrated in Fig. 11(b). It can be shown by trigonometry that

20

p: degrees

15

10

0 15

Fig. 9. p against dry for three different DS test device sizes with RMS graded sand

The test results show clearly that the size of the DS apparatus has a signicant effect on the determination of d 9s-p . The results from the 60 mm apparatus signicantly overestimate d 9s-p and p , compared with measurements with the larger devices, despite the fact that it complies with ASTM D3080-04 (ASTM, 2008d). The small device is also consistent with BSI 1377-7 (BSI, 1990), except that the ratio of box height to maximum particle size, H/Dmax, is about 9 compared with the recommended value of 10. Since most of the discrepancy between values of d 9s-p is related to differences in p , it appears that the 60 mm device tends to amplify dilatancy in the RMS graded sand. Given the systematically higher values of d 9s-p measured with the 60 mm conventional test apparatus, use of this

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

d/2 (dyy, dyx/2)

513
dsp

( )
N ds

psp

p ps R p 1 3 cos p 2

p psp cds cps 3

d

psp ps N (a) 1 3 2 1

psp

Pole for directions (a)

ps
(0, dxy /2)

psp c

1 3 2

Pole for planes ( yy, yx)

cps cps cotpsp

Nps 1 3 2 1 (b)

( ps, ps)

( N, )

Fig. 11. (a) DS and plane-strain strength parameters and their relationship with the top of the Mohr circle; (b) c and cps psp relationships
100
Plane-strain envelope at maximum obliquity psp 430 cps 12 kPa Linear fit: (1 3)/2 against (1 3)/2 343 c 09 kPa n 12

ds ps
O C

80

psp 430

60

: kPa

343 40

( xx, xy)

(b)

20

Fig. 10. Mohr circles for: (a) incremental strain; (b) stress (after Lings & Dietz, 2004)

1 3 1 3 c tan 2 2 1 3 1 3 cps cos ps-p sin ps-p 2 2

20

40

(7) (8)

60 N: kPa

80

100

120

(1 - 3)/2 against (1 3)/2 Linear fit: (1 3)/2 against (1 3)/2 Plane-strain envelope at max. obliquity

in which ps-p and cps are the plane-strain friction angle at peak DS strength and plane-strain cohesive intercept respectively. The radius of the Mohr circle, (1 3 )/2, is simultaneously satised by equations (7) and (8) if and only if c cps (9) cos ps-p ps-p sin1 tan (10)

Fig. 12. against N for partially saturated RMS graded sand at peak strength, dry 15.8 kN/m3

Hence equations (9) and (10) give the parameters that represent the plane-strain failure envelope from data plotted in accordance with equations (5) and (6). Figure 12 shows plots of (1 3 )/2 against (1 + 3 )/2

for partially saturated RMS graded sand with dry 15.8 kN/m3 at w 45% and for N from 2.1 to 98.9 kPa. A linear regression t to the data gives a line with a slope of 34.38 and an intercept of c 0.9 kPa. Using equations (9) and (10), the plane-strain envelope at maximum obliquity can be calculated from c and , and is also plotted in Fig. 12. This line has ps-p 43.08 and cps 1.2 kPa. As can be seen in the gure, the c envelope is a good t to the ((1 3 )/2, (1 + 3 )/2) data, and the ps-p and cps

514

OROURKE N was plotted in the gure. Linear regressions are shown for both datasets. Values of tan(p )p:sat: tan(p )dry from Fig. 13 are plotted in Fig. 14 against N . The differences in tan p are plotted as both the differences between the data points in Fig. 13 and the differences in the regression equations. The data point differences are tted with a power regression in Fig. 14, which has r 2 0.81. Also plotted in the gure is cds /N against N for cds 2.1 kPa, corresponding to the value measured for partially saturated RMS graded sand at dry 15.8 kN/m3 (see Fig. 12). Because of the form of the equations plotted in Fig. 14, it appears that cohesion increases without bound as N approaches zero. This is only apparent; the cohesion cannot exceed a maximum value related to the surface tension mobilised between soil particles. The three curves in Fig. 14 are statistically indistinguishable, showing that the difference in tan p for dry and partially saturated RMS sand is well represented by cds /N . The apparent cohesion generated in partially saturated RMS graded sand is therefore related to increased dilatancy. It appears that suction increases the interference among the sand particles, thus elevating the shear resistance of partially saturated sand compared with dry sand. This mechanism of cohesion related to dilatancy is consistent with critical-state concepts for the DS strength of sand described by Schoeld (2005).

envelope is a good approximation for the plane-strain envelope at maximum obliquity, as it passes closely to the tangent point of each Mohrs circle. As with ds-p and cds , ps-p and cps are assumed to be constant over N 2.198.9 kPa. COHESION IN PARTIALLY SATURATED SAND Taylors ow rule for the DS strength of dry sand can be expressed as     p 9 p tan d 9s-p N 9 dry N dry (11) sin c 9rit tanp dry and for the DS strength of partially saturated sand as   p cds tan ds-p N p:sat: N sin crit tanp p:sat: Data were obtained by Olson (2009) from DS tests on dry and partially saturated RMS graded sand with the same dry for N from 2.1 to 98.9 kPa. Moreover, all tests were prepared using the same method with the same applied energy. The partially saturated sand was prepared within a relatively narrow range of w 3.95.0%. Within the bounds of test measurement error, it was shown that there is no 9rit for dry signicant difference between tan d 9s-p sin c sand and tan ds-p sin crit for partially saturated RMS graded sand. Assuming these differences to be equal and combining equations (11) and (12) results in cds tanp p:sat tanp dry N (13)

(12)

which shows that the increase in dilatancy between dry and partially saturated sand is equal to the normalised cohesion, cds /N . Figure 13 shows tan p against N for dry and partially saturated RMS graded sand at dry 15.8 kN/m3 . When there was more than one test at a given N for either the dry or partially saturated sand, the average p of all tests at that
Partially saturated RMS graded sand Power fit: dry 158 kN/m3 1 ln y 058 ln x 012 r 2 095, n 7

10

MODELLING SOILPIPELINE INTERACTION As mentioned previously, to utilise the MohrCoulomb constitutive models in geotechnical software, it is necessary to know the plane-strain friction angle and cohesion for twodimensional (2D) analyses. The 100 mm DS test box fabricated in accordance with the design of Lings & Dietz (2004) was used to determine the appropriate strength parameters. The DS values of ds-d and cds were then converted to plane-strain values of ps-p and cps , following the procedures described above. To validate this approach, a series of nite element (FE) simulations were performed with the software ABAQUS (2006) using ps-p and cps determined for partially saturated RMS sand, and compared with the results of full-scale 2D tests performed at the Cornell NEES site. Fig. 15 presents a schematic diagram of the large-scale 2D test basin, which was lled with partially saturated RMS graded sand compacted in 200 mm lifts. Over 100 nuclear density gauge measurements of soil unit weight and w were made for each test. The basin was designed to measure the lateral force against displacement of pipelines through the application of
05 05 Difference in data point pairs cds/N, cds 21 kPa Power fit: difference in data point pairs ln y 095 ln x 035 n 7, r 2 081 Difference in regression equations Note: dry 158 kN/m3 for all data and regressions 02

tan(p)p.sat

01

01

tan(p)dry

Dry RMS graded sand Power fit: dry 158 kN/m3 ln y 019 ln x 234 r 2 075, n = 7 001 1 10 N and N: kPa Partially saturated RMS graded sand Linear fit: partially saturated RMS graded sand Dry RMS graded sand Linear fit: dry RMS graded sand 001 100

tan(p)p.sat. tan(p)dry

04

04

02

01

01

20

40 60 N and N: kPa

80

0 100

Fig. 13. tan(p )p:sat against N and tan(p )dry against N 9 for RMS graded sand at dry 15.8 kN/m3

Fig. 14. Values of tan(p )p:sat 2 tan(p )dry and c/N against N and N 9 for RMS graded sand at dry 15.8 kN/m3

cds /N

03

03

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Actuator 35 Hc/D 75 D 124 mm or 174 mm 22 Hbk/D 52 Interface elements

515

24 m

Hc

Reaction beam

24 m

Buried pipe 048 m D

Hbk

Direction of pipe movement

244 m

Fig. 16. Typical nite-element mesh for simulation of 2D largescale test results
Actuator (a)

Level of soil backfill

Direction of pipe movement Load cell Reaction beam Actuator 04 m Floor level (b) 12 m

the test basin and pipe circumference. The interface friction angle between the soil and pipe with HDPE external coating 1 and vertical basin wall (Formica and glass) was taken as 0:6d 9s-p on the basis of DS tests performed by Trautmann (1983) and Olson (2009). The interface friction angle between the soil and plywood base of the test basin was estimated as 0.8ds-p . A MohrCoulomb model was used in which ps-p 43.08 and cps 1.2 kPa, corresponding to dry 15.8 kN/m3 for partially saturated RMS graded sand. To represent strainsoftening, the model proposed by Anastasopoulos et al. (2007) was used to diminish linearly both ps-p and p to residual values at crit and 0 respectively, from the plastic strain at ps-p to the plastic strain at crit , using the results of DS testing as p f d xp d x y d xf d xp H d FE (14)

Fig. 15. (a) Plan and (b) cross-sectional elevation of full-scale 2D tests to measure horizontal force against displacement

horizontal force with the two long-stroke (1.2 m in one direction) hydraulic actuators as shown in the gure. Horizontal force was measured on each side of the box with a load cell, and lateral movement was measured with Temposonic displacement transducers. The loading arm was designed so that the test pipe could rise without vertical restraint as it was displaced laterally through the soil. The rate of pipe displacement was 2.4 mm/s. The test basin and loading conditions were similar to those used in previous full-scale tests (e.g. Trautmann & ORourke, 1985: ORourke et al., 2004), with the main exception being size. The internal dimensions of the test basin were 2.44 m 3 2.44 m in plan and 1.82 m in depth. The end effects of wall friction were minimised by the relatively large width of the test basin, and by lining the interior of the box with 1 Formica and glass, both of which provide for relatively low angles of interface friction. Tests were performed on pipelines 124 mm and 172 mm in external diameter, buried at a pipe centreline depth to diameter ratio Hc /D between 3.5 and 7.5. The test pipes had a 2.5 mm thick high-density polyethylene (HDPE) external coating, which is a typical coating for pipelines in the eld. Figure 16 shows typical geometric conditions incorporated in the FE model. The meshes were composed mostly of 700 to 800 eight-node plane-strain elements. A rened mesh was used within a distance of approximately two pipe diameters from the test pipe. Key dimensions are given in Fig. 16. The pipe steel had properties similar to those of ASTM Grade A-36 steel, and was assumed to be linear elastic. The test basin was xed in the x and y directions. Interface elements were used between the soil and both the interior surface of

in which p f is the shear strain increment beyond yield at which there is no dilation; H is the thickness of the DS specimen; dFE is the FE element size; and d xy, d xp, d xf are the DS test horizontal displacements at yield, peak strength and crit at which full softening occurs. In addition, cps was reduced from its maximum at p 9s-p to a small residual value of 0.1 kPa at p f . This reduction is consistent with measurements by Olson (2009) of little to no cohesion in DS tests for partially saturated RMS graded sand at large displacements. A FORTRAN subroutine developed by Robert & Soga (personal communication, 2009y ) to apply the Anastasopoulos et al. (2007) model in the MohrCoulomb model in ABAQUS was used in the 2D FE method simulations. Following the work of Yimsiri et al. (2004), laboratory test data were used to estimate the soil modulus for the 2D simulations. Multiple linear regression (MLR) analyses were performed on Youngs modulus E, vertical stress at pipe centre vc , and dry from the results of 25 full-scale tests conducted by Trautmann & ORourke (1985), Turner (2004) and Olson (2009) to obtain an expression for E as a function of vc and dry with highest statistical signicance. For partially saturated RMS graded sand an empirical equation was developed as  13:7 : 0:0378 E 2 3 1013 97 dry vc (15) in which dry and vc are expressed in units of kN/m3 and kPa respectively. Equation (15) provides an equivalent
Personal communication of FORTRAN subroutine for analysis of soilpipe interaction.

516

OROURKE test data for all Hc /D. In general, the FE results overpredict the measured dimensionless peak forces by a small margin of 18%. The favourable agreement between analytical and experimental results provides supporting evidence for obtaining ps-p and cps from DS data according to the approach described previously. It also marks a transition from characterising engineering behaviour through experimental results to characterising that behaviour on the basis of computational simulation. Over the past 25 years the experimental results of Trautmann & ORourke (1985) for dry sand have provided guidance in estimating lateral forces generated by soilpipeline interaction under large ground deformation (ASCE, 1984). All the experimental results of Trautmann & ORourke can now be replicated with delity by FE analyses. Moreover, the analytical methods are being extended to partially saturated sand, which is relevant for a much broader range of eld conditions. Soilstructure inter-

modulus that is empirically calibrated and suitable for direct use in the elasto-plastic MohrCoulomb model available in ABAQUS. Figure 17 shows the relationships of dimensionless force F9 against dimensionless displacement Y9 from the numerical simulations. The dimensionless force F9 F/t Hc DL, where F is the horizontal force, t is the total soil unit weight, and L is pipe length. The dimensionless displacement Y9 Y/D, where Y is the lateral displacement of the pipe with respect to the soil. The numerical simulations for partially saturated RMS graded sand show excellent agreement with the experimental data for all partially saturated tests. The numerical simulations compare favourably with the experimental results in terms of pre-peak, peak force, and post-peak behaviour. The FE and 2D test results for partially saturated RMS graded soil are compared in Fig. 18, in which dimensionless force Nq is plotted as a function of dimensionless depth Hc /D. There is excellent agreement among the analytical and
12

t 167 kN/m3 Pipe diameter 174 mm

12

t 167 kN/m3 Pipe diameter 124 mm

Dimensionless force, F/tHcDL

8 Hc

Dimensionless force, F/tHcDL

8 Hc

D 4

Olson 2D moist 3 Numerical analysis 0 0 02 04 06 0 0 02

Olson 2D moist 1 Numerical analysis

04

06

Dimensionless displacement, Y/D (a)

Dimensionless displacement, Y/D (b)

16

t 165 kN/m3 Pipe diameter 174 mm

16

t 165 kN/m3 Pipe diameter 120 mm

Dimensionless force, F/tHcDL

Dimensionless force, F/tHcDL

12

12

Hc 8

Hc 8

D 4 Olson 2D moist 6 Numerical analysis 0 0 02 04 06

D 4 Olson 2D moist 7 Numerical analysis 0 0 02 04 06

Dimensionless displacement, Y/D (c)

Dimensionless displacement, Y/D (d)

Fig. 17. Analytical and experimental dimensionless forcedisplacement relationships for partially saturated RMS graded sand: (a) Hc /D 3.5; (b) Hc /D 5.29; (c) Hc /D 6.5; (d) Hc /D 7.5

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

28
Maximum dimensionless force, Nq F/(HcDL)

517
N
Split box Pipe

10
Large-scale 2D partially saturated RMS graded sand tests RMS FEM simulation

Test 1: Axial Test 1: Bending Test 2: Axial Test 2: Bending

Test 3: Axial Test 3: Bending

24 20 16 12

8 6

Strain: %

4 2 0 Location of tactile force sensors

Springline

Crown

Hc

8 4 0 2 4 6 8
D Experimental data, r 075, n = 4 Trend of numerical simulation
2

2 5 4 3 2 1 0 1 Position: m S(), N()

Invert

Fig. 20. Comparison of strains in 400 mm diameter pipelines during large-scale tests

Dimensionless depth, Hc/D

Fig. 18. Dimensionless force against dimensionless depth for maximum lateral pipe force in partially saturated sand

action in partially saturated soils involves complex soil behaviour and modelling demands, and provides a future research area with great potential. HIGHLY DUCTILE PIPELINES High-density polyethylene (HDPE) pipelines represent a class of pipe with the capacity to accommodate a large amount of ground deformation associated with geohazards, such as earthquakes, ooding, subsidence and landslides. To investigate the capacity of HDPE pipelines for abrupt ground movement, and develop analytical capabilities for evaluating soilstructure interaction under such conditions, large-scale tests were conducted on HDPE pipelines subjected to 1.22 m of strike-slip displacement at a crossing angle of 658 with respect to ground rupture, as illustrated in Fig. 19. Each pipeline was instrumented with between 80 and 140 strain gauges, many with the capability of measuring strains as high as 20%. All tests were conducted with partially saturated sand, with the same dry , w, ds-p and cps described previously. The experimental pipelines consisted of HDPE pipes, manufactured by the Chevron Phillips Chemical Company under the commercial name DRISCOPLEX. Descriptions of the full-scale testing, measurement systems, and tests results are provided elsewhere (ORourke et al., 2008), and only select features of the test results are provided herein. The measured axial and bending strains along a nominal 400 mm diameter HDPE pipeline with wall thickness of 24 mm are plotted in Fig. 20 at a strike-slip displacement of

Final pipe position 32 m Initial pipe position

65

122 m Final box position Initial box position 106 m

Fig. 19. Plan view of large-scale pipeline test, with key dimensions and geometry

1.22 m. An inset diagram shows a schematic representation of the deformed shape of the pipe within the test basin. Another inset diagram shows the crown, invert and springline locations. Axial strains are the average of the pipe crown and invert strains, and bending strains are expressed as one half the difference between the springline strains. The bending strain so calculated is the incremental strain caused by pipeline exure relative to the axial strain. The maximum axial strains coincide with the ground rupture, and decrease with increasing distance from this location. Flexural strains are zero at the ground rupture, consistent with the point of counterexure for axisymmetric pipeline deformation. There is a remarkable degree of consistency in the measured bending strains. There are some differences in axial 1 strains, caused in part by Teon wrapping to protect special 1 sensors, as shown in the gure. In essence, the Teon isolates the pipeline from soil shear forces, resulting in higher axial strains at the south end of the pipe. The maximum measured strain was 8%, representing the combined axial and bending strains, and was located at a distance of approximately 1.0 m from the ground rupture either side of the fault. Moreover, the axial pipeline load decreased by 40% within 2 h after ground rupture. Because HDPE is viscoelastic, it has the benecial effect of reducing load with time at anchorages outside the ground rupture zone. The test results demonstrate the benets of HDPE pipeline ductility and viscoelastic response in accommodating permanent ground deformation. The maximum measured strains for 1.22 m of strike-slip displacement were far below strain levels associated with rupture or creep instability of the pipe wall. The tests, however, showed a 12% increase in vertical diameter at the location of ground rupture (ORourke & Bonneau, 2007), which was accompanied by 6% loss of the pipe internal cross-sectional area. The results therefore indicate that squeeze-off due to bending and locally high lateral soil forces is an important, potential failure mechanism for an HDPE pipeline with larger diameter-to-thickness ratio, D/t, affected by severe ground deformation. As illustrated in the next section, highly ductile pipelines can play a critically important role in the performance of water supplies vulnerable to earthquake-related ground deformation. The use of tests results at the macro scale provides information indispensable for improving system performance at substantially larger scales. Through study of the San Francisco water supply it is seen that system performance may be controlled by a limited number of critical pipelines. Improvements focused on strategically located components can therefore be used for effective system-wide risk reduction.

518

OROURKE and lifelines caused by permanent ground deformation (PGD). Investigations by Lawson et al. (1908), for example, identied areas of ll and made ground as zones of PGD and concentrated earthquake damage. The locations of water distribution pipeline damage were combined with observations of large ground deformation to delineate zones of inrm ground (Schussler, 1906; Manson, 1908) as a basis for planning future improvements in the water supply system. It is now known that much of the observed ground deformation was caused by soil liquefaction. Fig. 21(a) shows the zones of soil liquefaction in San Francisco mapped by Youd & Hoose (1978) from their study of historic ground failures in Northern California. The air photo used by Youd & Hoose has been modied to show the Marina, which was not shown in the original work. The Marina district was sparsely populated in 1906. Although permanent ground movements were observed there (e.g. Lawson et al., 1908), their effects on infrastructure were relatively minor. Fig. 21(b) shows a rendering of the zones of liquefaction-induced ground deformation effects developed with a geographical information system (GIS). Figure 22 is a map of the 1906 water supply within the city limits, developed from the oldest extant maps of the San Francisco water distribution system, reservoirs, and pressure districts (ORourke et al., 1992, 2006). There were nine reservoirs and storage tanks, with 92% of the total water storage contained in the Lake Honda, College Hill and University Mound reservoirs. The remaining reservoirs were low-volume units, which were drained within hours of the earthquake through ruptured and leaking water mains. All trunk pipelines, 400 mm or larger in diameter, are plotted relative to the zones of liquefaction-induced ground deformation delineated by Youd & Hoose (1978) and ORourke et al. (1992). Approximately 78% and 22% of the

SAN FRANCISCO The mainland city of San Francisco, CA, is located in an area of approximately 120 km2 , where its buildings and lifeline systems are a living laboratory for infrastructure response to geohazards at the kilo-scale level (see Table 1). With respect to earthquakes and infrastructure, San Francisco is distinctive for at least two reasons. First, it is the place where modern, geographically distributed systems were rst affected by a major earthquake. Second, it is a place where two severe earthquakes have occurred, generating liquefaction at the same locations and disrupting critical water supplies in similar ways. Fire following earthquake is regarded by many as the greatest single threat to San Francisco. Approximately 96% of the city buildings are timber structures (Scawthorn et al., 2006b), built in close proximity to each other. Fire in one such building will spread next door in about 10 min. With a 15 km/h wind, re unchecked will consume an entire block within 1 h, and spread to six adjacent blocks at conagration proportions in 2 h (Blackburn, personal communication, 2009{ ). San Francisco is a location where infrastructure and natural hazards converge. The presence of frequent high winds, fragile infrastructure and soils susceptible to liquefaction provides lessons from the 1906 San Francisco and 1989 Loma Prieta earthquakes important for infrastructure and geohazards worldwide. 1906 San Francisco earthquake The re following the 1906 San Francisco earthquake is the largest single re loss in US history (Scawthorn et al., 2006a), with approximately 490 city blocks burned to the ground, an additional 32 blocks severely damaged, and 28 000 buildings destroyed. The earthquake had a dramatic effect on civil infrastructure, including damage to buildings

1 km N

Marina Embarcadero Marina Approx. boundary of foot of Market Zone

Zones of soil liquefaction

Zones of soil liquefaction

U. S. Post Office

James Lick Skyway

Market St

Dore St

Approx. boundary of South of Market Zone Old Mission Bay

Valencia St

Approx. boundary of Mission Creek Zone Mission St

18th St

(a)

(b)

Fig. 21. Zones of historic liquefaction in San Francisco: (a) liquefaction zones (after Youd & Hoose, 1978); (b) liquefaction zones in GIS Personal communication with former Deputy Fire Chief of San Francisco Fire Department in command of reghters during 1989 Loma Prieta earthquake.

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Golden Gate Transmission pipelines Trunk pipelines ( 400 mm dia.) Zone of liquefactioninduced deformation Presidio Military Reservation Presidio Heights Reservoir Clay St Tank Francisco St Reservoir Lombard Reservoir

519

Bay ranc an F of S

ar ke tS

Pipeline break

8 Breaks

isco

Golden Gate Park 13 breaks

Lake Honda Reservoir

Clarendon Heights Reservoir

Valencia St

9 breaks Potrero Heights Reservoir

Scale

Pacific Ocean

Cond

uit

1000 m

College Hill Reservoir

citos

on

du

it

Fig. 22. Reservoirs, key pipelines and zones of liquefaction-induced ground deformation during 1906 San Francisco earthquake (ORourke et al., 1992)

trunk lines were composed of cast and wrought iron pipe, respectively. Breaks in the trunk lines crossing the liquefaction zones are plotted from the records of Schussler (1906) and Manson (1908). Multiple ruptures of pipelines from the College Hill and University Mound reservoirs occurred in the zones of liquefaction-induced ground deformation, cutting off nearly 60% of stored water to the city central business district (CBD). The ground deformation on Valencia St (see map location labelled 13 breaks) was perhaps the single most devastating event of the 1906 earthquake. It was responsible for the collapse of the Valencia St Hotel, killing 37 people, which was the second highest death count at an individual location during the earthquake (Fradkin, 2005). It also ruptured two cast iron water trunk lines, 400 and 550 mm in diameter. These broken pipelines emptied the College Hill Reservoir of 53 million litres, thereby depriving reghters of water for the burning CBD of San Francisco. The ground deformation at this location also destroyed a brick sewer, gas mains, electric and telephone conduits, and cable car tracks. Figure 23 is a map of the area that burned after the earthquake, relative to the water supply depicted in Fig. 22. The trunk lines of the College Hill and University Mound reservoirs north of the pipeline breaks are removed from this gure to demonstrate the lack of hydraulic capacity into the central city. The southernmost extent of the re was 20th Street, where water was still available from the intact pipeline network south of the liquefaction zone. The maximum western extent of the re coincides approximately with Van

Ness Avenue. Using this wide street as a rebreak, and drawing on water still available from Lake Honda, reghters were able to stop the western advance of the conagration. The computer program EPANET was used to model the ow in the undamaged trunk lines from Lake Honda. EPANET is a Windows-based software program distributed by the Environmental Protection Agency (EPA, 2008) for the simulation and analysis of ow and pressures in hydraulic distribution networks. In essence, the program solves a series of non-linear equations for continuity and energy conservation of incompressible ow in pressurised pipeline networks (Armando, 1987). Various parameters required to model the ow, including diameter, pipeline length, friction coefcient (C-value), and changes in elevation within the system were taken from existing maps and historic sources of information (e.g. Schussler, 1906; Manson, 1908), as well as digital elevation models for San Francisco. Figure 24 presents a bar chart showing the progressive loss of water in the Lake Honda Reservoir over four days from 7 am on the day of the main shock. The simulated reduction in reservoir level compares favourably with the successively lower levels actually measured and reported by Schussler (1906). The hydraulic network analyses show that total water ow from Lake Honda along Van Ness Avenue was approximately 20 00030 000 l/min, which is sufcient for effective action at a large rebreak such as Van Ness Avenue. The analyses also show that ows from Lake Honda were inadequate for re suppression within the burning

tal Cr ys gs Sprin uit d Con

Sa

nA

nd

Lake Merced

Pilar

rea

University Mound Reservoir

sC

520
Golden Gate Transmission pipelines Trunk pipelines ( 400 mm dia.) Zone of liquefactioninduced deformation Presidio Military Reservation Presidio Heights Reservoir

OROURKE
Francisco St Reservoir Lombard Reservoir Clay St Tank Burnt sections of San Francisco

of Bay

Golden Gate Park

cisc Fran San

N Lake Honda Reservoir

Clarendon Heights Reservoir

20th St

Potrero Heights Reservoir

Scale

Pacific Ocean

1000 m

College Hill Reservoir

ondu

it

os C

on

du

it

cit Pilar

Fig. 23. Reservoirs and key pipelines in relation to the re-damaged area of San Francisco in 1906 (ORourke et al., 1992)

120

100

Observation Simulation

80

60

40

20

0
7 am 18 Apr 1906 7 am 19 Apr 1906 7 am 7 am 20 Apr 21 Apr 1906 1906 Date and time 7 am 22 Apr 1906

Fig. 24. Comparison of simulated and observed water levels in the Lake Honda reservoir after the 1906 San Francisco earthquake

CBD, especially given the loss of supply from the College Hill and University Mound reservoirs. The 1906 earthquake underscores the critical importance of liquefaction hazards in San Francisco, the interaction between liquefaction-induced ground deformation and water distribution system performance, and the effects of water losses on catastrophic re and the destruction of the built environment. There is perhaps no better illustration of geohazard effects on critical infrastructure, nor a better demonstration of the importance of geotechnical factors for the design and management of large, geographically distributed systems. As described previously, highly ductile HDPE pipelines have been tested and evaluated for ground deformation at levels comparable to those sustained at Valencia Street and elsewhere in the 1906 zones of liquefaction-induced soil movement. Today, many cast iron pipelines are located in zones of potentially damaging liquefaction in cities at risk from earthquakes. In San Francisco, many of these zones are well known from previous earthquakes, and have been mapped to a high degree of accuracy with respect to location (e.g. ORourke & Pease, 1997; ORourke et al., 2006). Research and case history analysis show that the replacement of existing brittle pipelines (especially trunk lines) with wellconstructed HDPE or steel pipelines can substantially improve water supply performance during an earthquake. The opportunity for system improvement is especially keen in San Francisco, where critical water supply pipelines can be identied at known liquefaction hazards, and replacements prioritised for signicant impact on system performance.

Million litres of water

tal Cr ys gs Sprin uit d Con

Sa

nA

nd

Lake Merced

rea

University Mound Reservoir

sC

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS 1989 Loma Prieta earthquake After the earthquake and re of 1906 a supplementary water distribution network, known as the Auxiliary Water Supply System (AWSS), was constructed in San Francisco to provide emergency re protection. The AWSS has no building connections or service lines; only re hydrants can draw from the network. It is intended to augment the citys potable water distribution system, referred to as the Municipal Water Supply System (MWSS), by providing a system that works independently of, but in parallel with, the municipal supply. The AWSS is described elsewhere (e.g. ORourke & Pease, 1992; Scawthorn et al., 2006a), and only key features of the system are treated herein. At the time of the Loma Prieta earthquake it comprised 200 km of cast and ductile iron pipelines, with nominal diameters of 250500 mm. About 80% of the system was composed of cast iron pipelines. Earthquake damage was relatively low throughout the MWSS, with the exception of the Marina, where there were 123 repairs, caused principally by liquefaction-induced ground deformation (ORourke & Pease, 1992). Damage to the AWSS was light, but had serious consequences. Water ow through a 300 mm pipeline rupture at a location of soil liquefaction, supplemented by losses at broken hydrants, emptied a critical pressure-stabilising reservoir known as the Jones Street Tank. Loss of this reservoir led to the loss of water and pressure throughout the entire CBD of the city. In other words, the critical auxiliary supply that had been built and maintained after the 1906 earthquake for seismic protection was unable to supply water throughout large portions of San Francisco. Computer simulations of the AWSS network were performed with a special program. GISALLE (Graphical Interactive Serviceability Analysis for LifeLine Engineering), which had been developed as a rst-generation hydraulic network model to evaluate earthquake effects on large, geographically distributed water supplies (Trautmann et al., 1986; Khater et al., 1989). The computer model was built around a hydraulic pipeline network program that was modied to allow for simulation of post-earthquake damage states. A special code was developed and incorporated in the program to model the hydraulic performance of damaged pipeline systems with many breaks. The program accounted for the effects of ground deformation by modelling pipeline breaks in zones where soils are susceptible to liquefaction. Moreover, it was checked successfully against special re ow tests performed in the eld by the San Francisco Fire Department. Figure 25 shows a schematic plan view of the system that was simulated to reproduce the conditions on the night of the Loma Prieta earthquake. The AWSS is divided into a lower zone below an elevation of about 30 m and an upper zone above this level. The zones generally are isolated from each other by closed valves to prevent damage in the lower zone from depleting water resources at the higher elevations. The Twin Peaks Reservoir and Ashbury Tank supply the upper zone, and the Jones Street Tank supplies the lower zone. During a severe earthquake the Jones Street Tank is intended to provide elevated pressure and some ow for the lower zone, while one or both pump stations, shown in the gure, supply water from San Francisco Bay into the lower zone pipelines. Pump stations 1 and 2 were not activated immediately after the earthquake because of operator concerns and delay in communication, and are not included in the simulation. The locations of damage in the AWSS are indicated in the gure and identied in the legend. Figure 26 shows the results of the analysis in schematic format. Open arrows denote water egress either from the
FB Marina leak L Pump station no. 2 FB FB

521

Jones Street Tank H FB Pump station no. 1

H H H Ashbury Tank Folsom leak L

L H FB

Upper zone Lower zone Twin Peaks Leak Reservoir Hydrant break Pipe break Fire boat manifold

FB

Fig. 25. Schematic representation of the Auxiliary Water Supply System in San Francisco, and damage related to the 1989 Loma Prieta earthquake (ORourke & Pease, 1992)

Jones Street Tank or from damaged components. The solid arrows denote internal ow. Zones of potential soil liquefaction are denoted in the gure as the South of Market and Foot of Market areas. These zones are the same as those shown at similar locations in Fig. 21. The South of Market area was recognised as a zone of potentially unstable ground, called inrm ground when building the system (Manson, 1908), and was isolated from adjoining portions of the network by closed gate valves. Only one open gate valve was provided for the South of Market area at the intersection of Market and 6th Streets, as shown by the open circle in Fig. 26. This gate valve was designed to be operated remotely with utility-supplied electric power. Because of electric power loss at the time of earthquake, the valve could not be closed remotely. Consequently, water owed through this gate valve until the Jones Street Tank ran dry. The total ow rate from the Jones Street Tank was approximately 78 000 l/min. Given that the normal operating
Marina leak: 12 000

Jones Street Tank 78 000

29 000 10 000 13 000

14 000 7000 South of Market 38 000 15 000 4000 19 000 10 000 Folsom leak: 8000 Pipeline Hydrant break Pipeline break Isolation valve 6th and Market

Foot of Market

Zone of potential soil liquefaction Outflow: litres/min Internal flow: litres/min

Fig. 26. Hydraulic network analytical results for the Auxiliary Water Supply System during the 1989 Loma Prieta earthquake (ORourke & Pease, 1992)

522

OROURKE 1994 Northridge earthquake In 1994 Los Angeles was struck by the Northridge earthquake, with direct losses to the built environment exceeding $40 billion (Eguchi et al., 1996). Its impact on the water distribution system was substantial. Over three-quarters of the water supply for the City of Los Angeles was disrupted. Los Angeles Department of Water and Power (LADWP) and Metropolitan Water District (MWD) trunk lines (nominal pipe diameter > 600 mm) were damaged at 74 locations, and the LADWP distribution pipeline (nominal pipe diameter , 600 mm) system was repaired at 1013 locations (e.g. Jeon & ORourke, 2005). Next to the 1906 San Francisco earthquake, this event caused the most extensive earthquake damage to a US water supply system. Given the large earthquake-affected area, which coincides with the underground water distribution system and the widespread damage to that system, it was decided to treat the pipeline network as a giant strain gauge. Water distribution pipelines are constructed in a broad and relatively dense, rectilinear pattern. As a minimum, each pipeline acts as a binary extensometer, which provides site-specic information as to whether it was or was not repaired. The objective was to collect and normalise repair data so that seismic intensity could be visualised and relationships between damage and spatial variables quantied. The earthquake-induced damage to water pipelines and the database developed to characterise Northridge earthquake damage are described elsewhere (e.g. Toprak et al., 1999; Jeon & ORourke, 2005; ORourke & Bonneau, 2007), and only key features are presented here. The portion of the LADWP water distribution system most seriously affected by the Northridge earthquake is shown in Fig. 27. GIS databases for repair locations, characteristics of damaged pipe, and lengths of pipelines according to composition and size were assembled with ARC/INFO software. Nearly 11 000 km of distribution lines and over 1000 km of trunk lines were digitised and plotted with a geospatial precision of 10 m throughout the San Fernando Valley, Santa Monica Mountains, and Los Angeles Basin. This work was accomplished before a GIS system was developed by LADWP, and all digitisation was performed from analogue maps of the system. A properly calibrated strain gauge at mega scale requires a grid of repetitive structures having reasonably consistent properties and a damage threshold sensitive to the displacements being measured. Charts depicting the relative lengths of LADWP and MWD trunk and distribution lines according to pipe composition are presented in Fig. 28, where it is clear that cast iron (CI) pipelines were the most pervasive, repetitive structure in the Los Angeles distribution system at the time of the earthquake. The 7800 km of CI pipelines had the broadest geographic coverage, with sufcient density in all areas to qualify as an appropriate measurement grid. Moreover, CI is a brittle material subject to increased rates of damage at tensile strains of the order of 250500 microstrain, which are levels of ground strain attainable during strong seismic ground motion. The CI pipelines are therefore sufciently sensitive for monitoring variations in seismic disturbance. Figure 29 is a map of distribution pipeline repair locations and repair rate contours for CI pipeline damage. The repair rate contours, where repair rate is dened as the number of repairs per kilometre of pipeline, were developed by dividing the map into 2 km 3 2 km areas, determining the number of CI pipeline repairs in each area, and dividing the repairs by the distance of CI mains in that area. Contours were then drawn from the spatial distribution of repair rates, each of which was centred on its tributary area. A variety of grids were evaluated, and the 2 km by 2 km grid was found to

capacity of the Jones Street Tank is approximately 2.72 million litres, the time required to empty the Jones Street Tank would have been about 35 min. The estimated time to loss of tank agrees with observations during the earthquake. The computer simulation package also was used in planning studies before 1989 to predict correctly that the AWSS would not be reliable in a future severe earthquake. The results of coupled water supply and re simulations were presented to the San Francisco mayor and other city ofcials, and a bond measure was developed in 1987 for $46 million, which passed with 89% voter approval, to provide funding for rehabilitation of the AWSS and other re-related infrastructure (Scawthorn et al., 2006a). The bond approval provided the stimulus for developing a third system, called the Portable Water Supply System (PWSS), which consists of special vehicles, called hose tenders, carrying about 1.5 km of 125 mm diameter hoses and above-ground hydrants. The most serious re after the Loma Prieta earthquake broke out in the Marina district, where liquefaction-induced ground deformation severely damaged the MWSS, cutting off alternate sources of water. Although the Marina re could not be controlled with the conventional components of the water supply, it was brought under control by the PWSS. Three hose tenders were dispatched to the Marina, where they were able to contain and extinguish the re using water pumped from San Francisco Bay by the city reboat. The 1989 Loma Prieta earthquake once again shows the importance of liquefaction and permanent ground deformation for the fate of underground lifeline systems. Because of liquefaction, San Francisco came dangerously close to a second earthquake-related conagration. Fortunately, the effects of liquefaction-induced ground movement on system performance were anticipated. They were studied by means of computer simulations that led to a bond measure to improve the re department infrastructure, and a portable water supply that is able to draw from the inexhaustible source of San Francisco Bay. The earthquake demonstrates the interdependencies between lifeline systems. Failure of the electric power supply contributed to failure of the water supply because of interoperational links. The inability of the remotely controlled isolation valve of the AWSS to be activated by utilitysupplied electricity emphasises the importance of identifying critical interfaces between systems and providing suitable back-up for emergency conditions. The isolation valve is now connected to a battery pack that can be activated by radio transmission. The earthquake provides a good illustration of the benets of computer simulations of large, geographically distributed systems affected by geohazards. The AWSS computer simulations showed the spatial consequences of damage, providing the re department and city ofcials with the ability to visualise system-wide behaviour and take steps to improve performance before the next earthquake.

LOS ANGELES The City of Los Angeles water distribution system, which is operated by the Los Angeles Department of Water and Power (LADWP), is located in an area of approximately 1200 km2 . This area is roughly ten times that of mainland San Francisco at a size that qualies it at the mega-scale level (see Table 1). The LADWP system allows for examination of a large, geographically distributed system that is inuenced by large variations of geotechnical conditions, seismic response characteristics, and types and states of repair of pipelines and associated facilities.

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Foothill feeder LA aqueduct 1 and 2 5 I-118 Devonshire St 0 5 km

523

Balboa Blvd

I-5

I-2

10

Winnetka Av.

405

Roscoe Blvd

Interstate highways Main streets Trunk lines Distribution lines Santa Monica Bay

Topanga Canyon Blvd

State Hwy 134 Mulholland Dr.

on

ic

Bl vd

Sa

nt

I-10

n Ve

ice

d Blv

Fig. 27. Map of Los Angeles water supply system affected by Northridge earthquake (ORourke & Toprak, 1997)

Cast iron 11% Concrete 18%

Ductile iron 1% 10 000 Trunk lines : 1014 km Distribution lines : 10 750 km Concrete Riveted steel Cast iron Steel Asbestos cement Ductile iron 1000

Riveted steel 14% (a)

Steel 56%

Steel 11% Asbestos 9% Ductile iron 4%

(b)

Cast iron 76%

Length: km

100 LADWP LADWP (c) MWD

Fig. 28. Statistics for water trunk and distribution pipelines in the City of Los Angeles during the 1994 Northridge earthquake (ORourke & Toprak, 1997)

provide a good representation of damage patterns for the map scale of the gure (Toprak et al., 1999). By choosing the contour interval as the mean repair rate over the entire area, only those contours exceeding the system average are shown. The seismic intensity, calibrated against pipeline performance, correlates with the number of contour lines, with the

zones of highest seismic intensity identied as the areas of concentrated contours. In each instance, areas of concentrated contours correspond to zones where the geotechnical conditions are prone either to ground failure or to amplication of strong motion. Each zone of concentrated damage is labelled in Fig. 30 according to its principal geotechnical characteristics. In effect, therefore, Fig. 29 is a seismic

524

OROURKE
N

Santa Monica Bay Cast iron repair rate 13 06 Distribution pipe repairs Cast iron Steel Others Interstate highways Main streets

I-5

Topanga Canyon Blvd

Winnetka Av.

Balboa Blvd

Roscoe Blvd

I-405

Devonshire St

Mullholland Drive Contour interval 01 repairs/km

06

State Hwy 134

t an

ca ni

Bl

vd
I-10 05

d Blv ice n 5 Ve

10 km

Fig. 29. Cast iron pipeline repair rate contours for Northridge earthquake (ORourke & Toprak, 1997)

hazard map for the Los Angeles region, calibrated according to pipeline damage during the Northridge earthquake. Of special interest is the location of concentrated repair rate contours in the west central part of San Fernando Valley (designated in Fig. 30 as the area of soft clay deposits). Subsurface investigations reported by Holzer et al. (1999) found local deposits of soft, normally consolidated clay. Field vane shear tests disclosed clay with uncorrected vane shear undrained strength Suvst 2025 kPa at a depth of 5 m, just below the water table. Newmark sliding block analyses reported by ORourke (1998) provide evidence that near-source pulses of high acceleration were responsible for sliding and lurching on the soft, normally consolidated clay deposit. The principal cause of ground displacement and pipeline damage in the West Valley area is therefore shear deformation in the soft clay deposits. This phenomenon of permanent ground movement in the absence of liqueable soils is often referred to as lurching. As explained by Jeon & ORourke (2005), records from 164 strong-motion stations (screened from a larger dataset of 240 records) throughout the earthquake-affected area were collected and analysed with respect to various seismic parameters. Spatial distributions of the different seismic parameters were estimated by interpolation and superimposed on the pipeline network and spatially distributed database of pipeline damage. Using GIS software, the repair rate was calculated for areas inuenced by specic seismic parameters. Correlations were then developed through regression procedures to obtain the most statistically signicant relationships among repair rate, values of different seismic parameters and pipeline characteristics. Fig. 31 shows the CI pipeline repair rate contours superimposed on peak ground velocity (PGV) zones, which were developed by interpolating the maximum horizontal velocities recorded at the strong motion stations. Using the GIS database, a

pipeline repair rate was calculated for each PGV zone, and correlations were made between the repair rate and average PGV for each zone. As explained by ORourke (1998), similar correlations were investigated for pipeline damage relative to spatially distributed peak ground acceleration, spectral acceleration and velocity, Arias intensity, modied Mercalli intensity (MMI), and other indices of seismic response. By correlating damage with various seismic parameters, regressions were developed between repair rate and measures of seismic intensity. The most statistically signicant correlations for both distribution and trunk line repair rates were found for PGV. Such correlations are important for loss estimation analyses, which are employed to assess the potential damage during future earthquakes and develop corrective measures and emergency response procedures to reduce the projected losses (e.g. Whitman et al., 1997). Figure 32(a) presents the linear regression that was developed between CI pipeline repair rates and PGV on the basis of data from the Northridge and other US earthquakes. Fig. 32(b) shows repair rate correlations for steel, CI, ductile iron (DI) and asbestos cement (AC) distribution lines. The regressions indicate that the highest rate of damage for a given PGV was experienced by steel pipelines. This result seems surprising at rst, because steel pipelines are often substantially more ductile than CI or AC pipelines. Steel distribution pipelines in Los Angeles, however, carry the highest water pressures, and contain a subset of pipelines that are especially vulnerable to corrosion and slip at joints with gasket connections. Pipeline system performance and decision support As described by ORourke et al. (2008), system performance and modelling are important, for at least three main

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

525

Soft clay deposits susceptible to lateral 13 movement and lurching

Sands and interbedded clay/silts susceptible to liquefaction 06

I-5

Devonshire St

I-2

Topanga Canyon Blvd

Winnetka Av.

Balboa Blvd

10

Roscoe Blvd Steep slopes with soils and fills susceptible to slumping and landslides

Mullholland Drive

06

State Hwy 134

04

Saturated sand and clay deposits susceptible to site amplification

n Sa
Contour interval 01 repairs/km

ta

ca ni

bl

vd
I-10 05

lvd eB nic 5 e V

10 km

Fig. 30. Geotechnical characteristics of the areas of concentrated pipeline damage after the Northridge earthquake

13

06

Topanga Canyon Blvd

Winnetka Av.

Balboa Blvd

Roscoe Blvd

I-405

Devonshire St

Mullholland Drive Contour interval 01 repairs/km

06

St

Santa Monica Bay Cast iron repair rate Interstate highways Main streets Peak ground velocity: cm/s 010 1020 2030 3040 4050 5060 6070 7080 8090 90100 100170

15

Sa

a nt

on

ica

Bl

d Blv ice n 5 Ve

Fig. 31. Pipeline repair rate contours relative to Northridge earthquake peak ground velocity (ORourke & Toprak, 1997)

I-5
vd
05 0

04

I-10

10 km

526
4 3

OROURKE
4

Fit equation: log y 155 log x 815 R2 085

3 2

Fit equation (steel): log y 088 log x 528 R2 090

0100

Repair rate: number of repairs/km

010
9 8 7 6 5 4 3

9 8 7 6 5 4 3 2

Steel CI DI AC Fit equation (CI): log y 121 log x 678 R2 084 Fit equation (DI): log y 183 log x 939 R2 073 Fit equation (AC): log y 226 log x 1102 R2 071 10
2 3 4 5 6 7 8 9

Note: log ln, natural log 0010

9 8 7 6 5 4 3 2

1994 Northridge
2

1989 Loma Prieta 1987 Whittier Narrows 1971 San Fernando (south)

001
2 3 4 5 6 7 8 9

0001 10 PGV: cm/s (a) 100

PGV: cm/s (b)

100

Fig. 32. Pipeline repair rate correlation with PGV for: (a) cast iron water distribution lines; (b) steel, cast iron (CI), ductile iron (DI) and asbestos cement (AC) water distribution lines

reasons. First, system performance provides the basis for planning and engineering at a scale commensurate with earthquake or other hazards that have large, geographically distributed effects. Second, system performance is the logical extension of component and/or individual pipeline response. It entails the outcome of integrated component behaviour, and for a pipeline network represents the ultimate expression in terms of service and the consequences of soilstructure interaction. Third, system performance provides the only way by which managers and engineers can gauge the scale and regional impact of an earthquake or similar natural hazard. System performance sets the stage for quantifying the regional economic consequences and community impact of an earthquake, as well as planning for emergency response and system restoration. The current generation of hydraulic network models for large, geographically distributed water supplies has evolved sufciently that engineers and managers can use them to plan and design for complex performance under the highly variable and uncertain conditions associated with geohazards. Simulations are run for a suite of different scenarios that allow system personnel to visualise a wide range of responses for an entire system or a specic part of that system. By running multiple scenarios, with and without modications of the system, engineers and managers can identify recurrent patterns of response and develop an overview of potential performance, helping them plan for many eventualities and improving their ability to improvise and innovate during an extreme event. The plan that emerges from any particular suite of scenarios, however, is not as important as the planning process itself, because as soon as a disaster unfolds, the reality of the event will diverge from the features of the most meticulously designed scenario. With good planning, however, emergency managers and lifeline operators can improvise, and skilled improvisation enables emergency responders to adapt to eld conditions. A decision support system is a computer-based information and modelling system that works interactively with users to address unstructured problems for strategic plan-

ning, management and operations (Turban, 1995). Such a system was developed for water supplies by using LADWP as a test bed. The system is intended to plan operations, emergency response, and new system facilities and congurations to optimise water supply performance during and after earthquakes (ORourke et al., 2008). It is generic, and the architecture of its computer programs is adaptable to any water supply. The system works in conjunction with the previously described hydraulic network model, EPANET, which is available on-line from the US Environmental Protection Agency (EPA, 2008), as well as a special program for damaged network ow modelling, known as Graphical Iterative Response Analysis for Flow Following Earthquakes (GIRAFFE). Detailed information about the development and evaluation of GIRAFFE is provided by Bonneau & ORourke (2009). The decision support system uses a hydraulic network model that accounts for all 11 700 km of water trunk and distribution pipelines and related facilities (e.g. tanks, reservoirs, pressure regulation stations) in the LADWP system. The system also accounts for the aggregated seismic hazard in Los Angeles through an ensemble of 59 scenario earthquakes. The 59 scenario earthquakes provide a library of seismic scenarios, from which engineers can select specic scenarios or combinations of scenarios to assess system performance. The decision support system works with risk and reliability assessment tools to provide metrics of system performance. The computer simulations account for the interaction of the water and electric power supplies, and model output can be used to evaluate the regional economic and community impacts of water losses. All system input and output can be visualised through GIS with advanced query logic and web-based features. The simulations are dynamic in time, and can account for loss of service as tanks and local reservoirs lose water over time through leaks and breaks in pipelines. As explained by ORourke et al. (2008), the LADWP decision support system was validated through comparison of simulation results showing system response to the 1994

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS Northridge earthquake with the actual areas of lost water service, as well as pre- and post-earthquake measurements of ow documented by LADWP after the Northridge earthquake. There is very good agreement between model results and LADWP records with respect to system-wide serviceability, geographic distribution of lost service, and pre- and post-earthquake ows over time at key locations. Pipeline damage caused by permanent ground deformation (PGD) is accounted for explicitly by locating the area of the system subject to large ground movements and estimating the damage by various methods, including expert judgement, simplied or site-specic FE models for soilpipeline interaction, and fragility curves for pipeline response. Pipeline damage caused by transient ground deformation (TGD), or seismic waves, is estimated by means of regressions developed from previous earthquake records. Regressions developed by Jeon & ORourke (2005) and Wang & ORourke (2008), for water distribution and trunk lines respectively, are used. Because it is not possible a priori to know the exact damage locations, multiple system response analyses known as Monte Carlo simulations are run, and the statistics of the simulated performances are summarised. As explained by Bonneau & ORourke (2009), Monte Carlo simulations are run according to an algorithm that uses a Poisson process to simulate the occurrence of pipeline damage. The user can specify the number of Monte Carlo simulations, or allow the program to determine the number of simulations when convergence criteria have been met. Typically, 15

527

simulations were required to meet the convergence requirements for the simulations discussed in this paper. In addition to modelling pipeline damage, it is important to account for the vulnerability of other facilities. For example, tank damage was modelled by fragility curves developed for different types of tank used by LADWP. The simulations incorporate fragility curves proposed by ORourke & So (2000) for steel tanks as well as fragility curves for concrete tanks used in the loss estimation program HAZUS (FEMA, 2006). Scenario earthquakes System simulations were performed for the Los Angeles water supply response to a 7.8 MW earthquake on the southern San Andreas fault (SAF), known as the ShakeOut scenario earthquake (USGS, 2008). This scenario was used as part of an earthquake preparedness exercise referred as the Great Southern California ShakeOut, which was the largest earthquake preparedness drill in US history, with an estimated 5.5 million people participating. Figure 33 shows the strong motions predicted by the ShakeOut scenario throughout the LADWP system. Fig. 33(a) shows a regional map of the Los Angeles area in which the LADWP system is located, relative to the SAF. Fig. 33(b) presents the spatial distribution of PGA generated by the earthquake. The San Fernando Valley and the southern part of the service area are the locations of maximum

SAF

N SAF N

LADWP system (a) N

0 10 km

PGV and fracture PGV: cm/s 020 2040 4060 6080 80100 100120 120140 140160 160180 180200 200500 0 (c) 10 km

PGA PGA: g 000010 010015 015020 020025 025030 030035 0 (b) 10 km

Fig. 33. (a) Location of San Andreas fault (SAF) and LADWP system; (b) PGA contours; (c) PGV contours and SAF (Romero et al., 2010)

528

OROURKE The system serviceability index (SSI) is the same percentage for the entire system. All simulations discussed here were run for the average 24 h summer daily water demand. Permanent ground movements predicted with simplied models in the ShakeOut scenario (USGS, 2008) were used to evaluate PGD damage to water supply pipelines, using a decision process developed with LADWP that accounts for pipe type and the predicted magnitude of movement (Romero et al., 2010). Figure 35 shows the ow conditions in the trunk line system at 0 and 24 h after the earthquake for a single network analysis representing the median results of the Monte Carlo simulations. This gure provides information about the spatial distribution of ows, and the way they diminish with time. The deterioration in performance is generated by losses from leaking pipelines that draw down tanks and reservoirs, causing some sections of the system to lose all local sources of water. Following such a large event, it will take a considerable time to isolate and repair leaking pipelines. A period of 24 h was chosen in consultation with LADWP personnel as a representative interval for showing time-dependent losses before signicant repair and restoration can be initiated. The decrease in pipelines with reliable water ow and the increase in unsatised demand nodes are clearly shown by comparison of the 0 and 24 h conditions. The mean SSI declines from 76% to 34% over 24 h, which indicates that 66% of the normal water demand cannot be met one day after the main shock. Figure 36 provides histograms of outcomes for the LAWDP system response to the ShakeOut earthquake scenario with and without the disconnected reservoirs. Because it is not possible to know a priori where pipelines will be damaged by ground wave effects, the simulated earthquake performance outcomes must be sampled until a statistically signicant number is acquired, as previously explained. The SSI statistics for 24 h after the earthquake are summarised in the gure, where the number of Monte Carlo simulations that contribute to a particular SSI was divided by the total number of simulations to provide an approximate probability index. The histograms of probability allow one to compare performance outcomes when the Encino, Lower Stone Can-

PGA (approximately 0.3g). Fig. 33(c) presents the SAF rupture and distribution of PGV. The locations of maximum PGV correspond to deep sediment basins, which amplify the incoming ground waves to produce locally high PGV. There are two main locations of wave amplication in the northern part of the San Fernando Valley and the southeast part of the region, where PGVs approach 200 cm/s.

Application of decision support system The LADWP is undertaking an extensive capital improvement programme to meet the requirements of the US Environmental Protection Agency and California State Department of Health Services with respect to surface water treatment and disinfection by-products. Signicant water system changes are necessary to meet the requirements. System changes include the removal of Encino, Hollywood and Lower Stone Canyon reservoirs from normal operating service, which places a much greater importance on the Los Angeles Reservoir and Van Norman Complex for reliable water distribution. Fig. 34(a) shows the locations and approximate water storage capacities of the Los Angeles, Encino, Lower Stone Canyon and Hollywood reservoirs. The removal of these reservoirs represents a loss of about 34 million m3 of water from immediate use in the system. The system response was evaluated for 15 water service areas, shown in Fig. 34(b). Water service areas are geographic groupings of pipelines, pumps, valves, tanks, reservoirs and demands that can be analysed individually. From north to south the water service areas are: Granada Hills (GH): Foothills (FH); Sunland-Tujunga (ST); Valley Floor A, B and C (VF A, VF B, VF C); Encino Hills (EH); Santa Monica (SM); Hollywood Hills (HH); Mount Washington (MW); Highland Park (HP); Santa Ynez (SY); Westside (WS); Central City (CC); and Harbor (H). By showing the results for the 15 water service areas, it is possible to understand the spatial variability of the system performance as expressed in terms of serviceability index (SI), which is the percentage of post-earthquake ows relative to pre-earthquake ows at all demand nodes within a water service area.

LA reservoir 136 M m3

GH VFB

FH ST

Encino Reservoir 148 M m3 Lower Stone Canyon Reservoir 136 M m3

VFC Hollywood Reservoirs 54 M m

3

VFA EH SM WS CC HH MW SY HP

10 km

10 km

(a)

(b)

Fig. 34. Maps of Los Angeles water supply showing: (a) reservoirs; (b) water service areas

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

N N

529

Flows and demands Pipes Damaged Flow Non-functional Unsatisfied demands 0 10 km 0 10 km

(a)

(b)

Fig. 35. Simulated water ows and water demands, (a) immediately after and (b) 24 h later for the MW 7.8 ShakeOut earthquake scenario (Romero et al., 2010)

04
Reservoirs open Reservoirs closed

025
Reservoirs open Reservoirs closed

03

Probability

02 Reservoirs open Mean: 0417 Median: 0396 St dev: 0044 Min: 0365 Max: 0509

Probability

Reservoirs closed Mean: 0343 Median: 0347 St dev: 0027 Min: 0280 Max: 0371

020

Reservoirs closed Mean: 0213 Median: 0214 StDev: 0062 Min: 0062 Max: 0356 Reservoirs open Mean: 0383 Median: 0385 StDev: 0089 Min: 0249 Max: 0581

015

010

01

005

0 0 005 010 015 020 025 030 035 040 045 050 055 060 Serviceability index

0 0 005 010 015 020 025 030 035 040 045 050 055 060 Serviceability index

Fig. 36. Histograms of system serviceability indices with and without disconnected reservoirs for the MW 7.8 ShakeOut earthquake scenario

Fig. 37. Histograms of serviceability indices for the most populated water service areas with and without disconnected reservoirs for the MW 7.8 ShakeOut earthquake scenario

yon and Hollywood reservoirs are closed and open. The median SSI increases by about 6% when the reservoirs are open as opposed to closed. The shift in the probability distributions can be seen clearly in the gure. Figure 37 provides a similar display in which service area SIs associated with the most populated areas of Los Angeles are represented. These areas include water service areas WS, CC, HP and MW, which are geographically close to the Encino, Lower Stone Canyon and Hollywood reservoirs.

Because of their large populations, these areas of Los Angeles are likely to have the greatest need for water to ght post-earthquake res. The probability distributions for the most populated areas are substantially different from those for the entire system. The median SI decreases from 38% with reservoirs open to 21% with reservoirs closed. Perhaps the most important nding is that the worst-case outcomes with reservoirs closed vary from SI of 5% to 15%. Such low levels of water

530

OROURKE Figure 38(b) shows the results 24 h after the Northridge earthquake for TGD-induced pipeline damage. The gure presents the same information as in Fig. 38(a), except that only TGD-induced damage is presented. The SSI immediately after the earthquake is 100%, falling to 75% after 24 h. The TGD-induced damage reduces the ow in the northern and southern parts of the San Fernando Valley relative to the ow for PGD-induced damage. Figure 38(c) shows the system performance 24 h after the Northridge earthquake for combined PGD- and TGDinduced pipeline damage. The SSI immediately after the earthquake is about 93%, falling to 74.6% after 24 h. Combining the PGD and TGD damage produces a 24 h SSI that is only 0.4% lower than that associated with TGD damage only. A careful examination of the simulation results discloses that the TGD damage is widespread, and affects important trunk lines that are old and vulnerable to seismic wave effects. In contrast, the PGD-induced damage is localised, being concentrated at a few specic areas of liquefaction, and lurching in the northern part of San Fernando Valley. When the PGD-induced damage is added to the TGD-induced damage, the hydraulic capabilities of the damaged system are not signicantly changed. For a complex system such as the Los Angeles water supply the network performance is highly non-linear. Adding PGDinduced damage, which produces an 11% decline in serviceability over 24 h, to TGD-induced damage does not result in a proportionate drop in system performance. The results show a markedly different system response in Los Angeles from the system response to earthquakes in San Francisco. The San Francisco system is smaller, and lacks sufcient dispersion and redundancy to compensate for liquefaction-induced damage to key pipelines. The fate of the system is truly the fate of the ground, primarily with respect to large deformation and failures that can incapacitate the network. In contrast, the Los Angeles water distribution system is much larger, with alternative paths for ow should localised PGD disrupt a limited number of pipelines and associated facilities. Moreover, there is vulnerability in Los Angeles to TGD effects, where the widespread passage of ground waves seeks out the most vulnerable parts of the system.
N

service would expose the most populated areas of Los Angeles to exceptionally high risk. With reservoirs open, the probability distributions shift markedly to the right, such that the worst-case scenarios have SIs higher than the mean SI for reservoirs closed. In some cases water is available at nearly 50% of the demand nodes after 24 h. The results of the scenario earthquake simulations show that opening the disconnected reservoirs immediately after a serious earthquake improves serviceability signicantly in the locations of highest population, and is thus an effective means of emergency response, even though such action will require tap water safety notices to be issued for the entire system. Work, such as this, that links emergency response with public health decisions provides a good example of how system modelling integrates geotechnical effects on a regional scale with the complex performance of a critical lifeline network to make difcult decisions involving public health and emergency response. Permanent versus transient ground deformation effects The hydraulic network model for the LADWP system provides an opportunity to evaluate how the Los Angeles water supply was affected by PGD and TGD during the 1994 Northridge earthquake. Because Northridge earthquake damage to the LADWP pipelines has been carefully documented with respect to PGD and TGD effects (Bonneau & ORourke, 2009), it is possible using the LADWP hydraulic network model to simulate system response to both the independent and combined effects of PGD and TGD. Figure 38(a) presents the results of the simulations performed with only PGD-induced damage 24 h after the Northridge earthquake. This gure shows the locations of PGD-induced damage, pipelines with and without ow, and pipelines not normally connected but available for emergency operations. The gure also shows PGV contours interpolated from 164 strong ground motion records, which are the same as those illustrated in Fig. 31. The GIS layer containing the ocean and coastal boundary is not shown so that the pipeline network and strong motion patterns are more clearly visualised. The SSI immediately after the earthquake is about 98%, falling to nearly 89% after 24 h.

PGV: cm/s 0 0 10 km

160

PGV: cm/s 0 0
Pipes with flow

160

PGV: cm/s 0 0 10 km

160

10 km

PGD damage Pipes with no flow

Pipes not connected

TGD damage Pipes with flow

Pipes with no flow Pipes not connected

TGD damage PGD damage

Pipes with flow Pipes with no flow Pipes not connected

(a)

(b)

(c)

Fig. 38. Simulation results for Los Angeles water supply damage at 24 h after the Northridge earthquake: (a) PGD only; (b) TGD only; (c) PGD and TGD

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS Scale must be remembered, of course. Localised ground failure in Los Angeles can reduce local serviceability without a dramatic reduction in system serviceability. Because of its size, however, local reduction of serviceability in Los Angeles can affect areas of the system comparable in size to that covered by the entire water distribution network in San Francisco. HURRICANE KATRINA Hurricane Katrina is generally recognised as the greatest US natural disaster. As previously indicated, its direct cost is over $100 billion (Jordan & Paulius, 2006), of which nearly $25 billion are direct property losses in New Orleans (IPET, 2008a). Over 80% of New Orleans was ooded, in some places to depths of 5 m. It took 53 days to dewater the city. Residential property losses were approximately 78% (IPET, 2008a), resulting in the displacement of hundreds of thousands of people, many of whom have not returned. In 2006 and 2008 (one and three years after Katrina) the population of New Orleans was only 43.5% and 64.3%, respectively, of its pre-hurricane level (US Census Bureau, 2009). There are nearly 2000 dead and missing attributed to Hurricane Katrina. Hurricane Katrina also had serious consequences for Gulf of Mexico oil and natural gas production. After Hurricane Katrina, 100% of oil and 94% of gas production in the Gulf of Mexico were suspended (MMS, 2006). Restoration of production was interrupted 26 days later in response to Hurricane Rita, which in turn resulted in the suspension of 100% of oil and 80% of gas production. Hurricanes Katrina and Rita destroyed 111 offshore platforms, with serious damage to 52 others, and severely damaged 169 submarine pipelines (Wisch & Ward, 2007). Predicting mudslide damage to submarine pipelines is an ongoing concern. Risk analyses link mudslides with hurricane-induced wave forces, and show the geographical distribution of potential pipeline and offshore platform damage based on mudslide recurrence intervals (Nodine et al., 2007). Much has been written about Hurricane Katrina and its implications with respect to geohazards and geotechnical engineering. The most comprehensive study of the hurricane effects on the New Orleans and southeast Louisiana HPS was prepared by the Interagency Performance Evaluation Task Force (IPET, 2008a). It comprises over 7500 pages, covering geodetic vertical and water level measurements, HPS characteristics and performance during Hurricane Katrina, characteristics of the hurricane, consequences of the disaster, and risk and reliability analyses. A special issue of the ASCE Journal of Geotechnical and Geoenvironmental Engineering (Christian, 2008) was published, with 15 papers covering most geotechnical aspects of the hurricane. Other notable reports include, but are not conned to, those by the National Academy of Engineering and National Research Council (NAE/NRC, 2009), the ASCE Hurricane Katrina External Review Panel (2007), and the Independent Levee Investigation Team (2006). It is the intention of this paper to provide a brief summary of Hurricane Katrinas effects and the performance of the HPS. Only select features are covered. Reference should be made to the publications identied above for more detailed information and analyses. Hurricane protection system (HPS) As shown in Fig. 39, the HPS consists of approximately 565 km of protective structures, including earthen levees and oodwalls, surrounding primarily ve counties, or parishes, adjacent to the Mississippi River for a total protected area of 5400 km2 . The system is composed of three hurricane
Lake Pontchartrain Lake Pontchartrain and vicinity Jefferson New Orleans Lake Borgne

531

West Bank and vicinity Mississippi River Plaquemines St. Bernard

New Orleans to Venice Chandeleur Sound Federal HPS State and local levees Barataria Bay Gulf of Mexico 0 km 10 N

Fig. 39. Hurricane protection system for New Orleans and southern Louisiana (after IPET, 2008a)

protection projects, identied in the gure as the Lake Pontchartrain and vicinity, West Bank and vicinity, and New Orleans to Venice Projects. The levees and oodwalls provided life safety protection for approximately 1.1 million pre-Katrina residents of metropolitan New Orleans, as well as security for one of the most important commercial and industrial hubs in the Gulf of Mexico region. By any interpretation of the term the HPS is a lifeline, crucial for the economic and societal well-being of a major port and urban centre. The HPS was authorised by the US Congress under the 1965 Flood Control Act after Hurricane Betsy caused catastrophic ooding in New Orleans. Twenty-seven years later, after multiple design changes, environmental challenges in the courts, and continuous interaction with local communities, a nal resolution was reached on how to build the system. When Hurricane Katrina struck New Orleans the HPS was incomplete, with no parish having the full level of protection authorised in 1965. As emphasised by IPET (2008a) and the NAE/NRC review (NAE/NRC, 2009), the HPS did not perform as a system. It was constructed in a piecemeal fashion over many years that represented a history of continuous incompleteness (IPET, 2008a). Figure 40 is an aerial view of the main Orleans East Bank Metropolitan Basin, which is the most populous area of New Orleans, containing the main downtown area and historic French Quarter. It is surrounded by a portion of the HPS, and is drained by three outfall canals (17th Street, Orleans and London Avenue canals), which convey water pumped from the basin to Lake Pontchartrain, immediately north of the city. As part of the resolution reached with local communities in 1992, a parallel protection system of levee and I-wall structures was built along the outfall canals as the sole measure to contain hurricane surge from Lake Pontchartrain. An I-wall uses a steel sheetpile driven into an existing levee to increase the height of the ood protection structure, and thus increase its capacity to resist a higher level of storm surge. Fig. 40 shows the Inner Harbor Navigational Canal (IHNC), which connects Lake Pontchartrain, the Mississippi River, and a canal, which at the time of the hurricane served as a channel for both the Gulf Intercoastal Waterway and the Mississippi River Gulf Outlet. As described in detail elsewhere (e.g. IPET, 2008a; Seed et al., 2008b), that canal acted as a conduit for storm surge, elevating waterway levels and breaching the HPS at several locations along the IHNC. A review of the regional geology (Dunbar & Britsch, 2008; IPET, 2008a; Rogers et al., 2008) discloses that much of the soil within 1020 m of the ground surface is composed of organics and normally to lightly consolidated

532

OROURKE
Lake Pontchartrain North London Avenue canal breach 17th Street canal breach

Inner Harbor Navigational Canal

South London Avenue canal breach

Orleans Canal Inner Harbour navigation canal breaches

Gulf Intercoastal Waterway Lake Pontchartrain

Area of interest New Orleans CBD 0 km Lake Borgne 10

Mississip

pi River

0 km

2 N

Fig. 40. Aerial view of main Orleans East Bank metropolitan basin (after Sasanakul et al., 2008)

clays. Subsidence triggered by drainage, atmospheric exposure of organics to decomposition, groundwater withdrawal and placement of lls altered the datum of many city benchmarks, with the result that some oodwalls were constructed approximately 0.6 m below the levels intended by design. The HPS was designed for a standard project hurricane that was developed from meteorological data acquired from 1900 to 1956. This design approach underestimated the storm surge generated by Hurricane Katrina, which was measured as 56 m on the eastern side of New Orleans compared with the 3.74.4 m design levels associated with this area. The HPS had not been designed for protection against overtopping, with no armouring of the back sides of the levees. Moreover, some levees were constructed with erodible soils that were dredged from local sources, thus leading to substantial erosion and breaches in the system. According to IPET (2008a) there were 50 major breaches in the HPS, four of which are attributed to foundationinduced failures. Three of these failures occurred along the outfall canals to Lake Pontchartrain (one at the 17th Street Canal and two at the London Avenue Canal), with the remaining one located on the east side of the IHNC near its intersection with the eastwest shipping channel. There is disagreement between IPET (2008a) and the Independent Levee Investigation Team (2006), which claims that a fth breach, further south along the east side of the IHNC, was most likely caused by loss of soil strength due to underseepage and consequent lateral instability (Seed et al., 2008b). Both IPET and the Independent Levee Investigation Team note that many of the breaches occurred at transitions, where weaknesses existed at connections between ood walls and earthen levees, as well as at penetrations of the HPS by pipelines, rail lines, roads etc. (IPET, 2008a; Seed et al., 2008b). The HPS along the outfall canals to Lake Pontchartrain

was designed for a safety factor of 1.3, and the soil strengths assumed in design were higher than warranted (IPET, 2008a). Attention has been drawn to the inadequacy of such a low safety factor, as well as its inconsistency with the higher safety factors used routinely in the design of earth dams affecting populated areas (e.g. ASCE Hurricane Katrina External Review Panel, 2007; NAE/NRC, 2009). Of critical concern are the reasons for the foundationinduced failures of the HPS. Centrifuge studies were performed that demonstrate the loading mechanism involved in the failures (Sasanakul et al., 2008; Ubilla et al., 2008). The photograph in Fig. 41(a) shows the results of a centrifuge test simulating the conditions associated with the 17th Street Canal failure. The hydrostatic force associated with rising water causes a gap to open between the sheetpile and soil on the outboard side of the I-wall. Water ows into the gap, generating hydrostatic pressure that extends to the base of the sheetpile, and a sliding surface develops from the toe of the pile into the landward side of the I-wall. Fig. 41(b) provides an illustration of this loading mechanism (IPET, 2008a). There is general agreement within the engineering community that the loading mechanism described above contributed to the I-wall failures at the 17th Street, London Avenue and Inner Harbor navigational canals. There is disagreement, however, with respect to soil failure on the inboard side of the 17th Street canal I-wall. Investigations and analyses by the IPET team (Brandon et al., 2008; Dunbar and Britsch, 2008; Duncan et al., 2008; IPET, 2008a) indicate that the most likely mechanism for soil failure was sliding in lacustrine clay, just beneath its interface with overlying swamp/ marsh deposits that were immediately below the 17th Street canal I-wall. Investigations and analyses by the Independent Levee Investigation Team (2006) indicate that the most likely mechanism for soil failure was sliding along a thin

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Sheetpile wall Water penetration behind wall Clay Organics Clay Organics

533

Clay

Sliding in clay (a)

C L

Deflection and pressure

Failure and movement C L (b)

Fig. 41. (a) Centrifuge test cross-section and (b) loading mechanism for New Orleans I-walls (IPET, 2008a; Sasanakul et al., 2008)

(2030 mm) layer of sensitive organic clay at a higher elevation in the soil prole (Seed et al., 2008c). The sensitive clay layer was located between the swamp/marsh deposits and an underlying layer of soft clay and silt lenses that has been interpreted as soil derived from the intermixing of salt and freshwater environments (Rogers et al., 2008). The lack of consensus on the mechanism for soil failure at the 17th Street canal represents a signicant gap in the lessons derived for geotechnical engineering from Hurricane Katrina. Steps have been taken to limit problems with Iwalls in future hurricanes, and thus the urgency for resolving the mechanism of soil failure has been reduced. I-walls along the outfall canals have been isolated from storm surge effects by the construction of ood gates and pump stations at the canal junctions with Lake Pontchartrain. Elsewhere, many I-walls have been replaced by stronger ood walls. Nonetheless, uncertainty regarding the 17th Street canal foundation failure should be resolved. The NAE/NRC review (NAE/NRC, 2009) concluded that the IPET explanation for the failure mechanism at the 17th Street canal while plausible, is not fully convincing, and alternative failure mechanisms should be more rigorously assessed. The NAE/ NRC review further encouraged that consideration in the design of levees and oodwalls be extended to all reasonable, possible failure modes. Repairs and strengthening of the HPS are being implemented in a massive construction programme led by the US Army Corps of Engineers (USACE) to create a new hurricane and storm damage risk reduction system (HSDRRS) by 2011. It will provide protection against a hurricane with a 100 year return interval (1% annual probability of failure). The standard of a 100 year hurricane is derived from the US National Flood Insurance Program, which sets ood insurance requirements for buildings below the 100 year ood

elevation (e.g. Burby, 2001). Building the HPS for a 100 year ood excludes protected property from designation as a ood hazard area according to federal insurance requirements, and is used frequently as a de facto design standard in US oodplains. The 100 year ood level will result in an HPS signicantly more resilient than the pre-Katrina HPS. As of 2008, reconstruction had produced an HSDRRS that performed at a level substantially better than what would have been expected from the pre-Katrina HPS during Hurricane Gustav, which was a Category 2 hurricane generating storm surge in the IHNC less than 1 m below Hurricane Katrina levels (Wooten et al., 2009). Nevertheless, questions have been raised about a 100 year ood or hurricane standard for heavily populated areas (e.g. ASCE Hurricane Katrina External Review Panel, 2007; NAE/NRC, 2009). Such a standard is low compared with recurrence intervals adopted, for example, in the Netherlands for ood protection systems, which use 400010 000 year return intervals (Hoeksema, 2006). It has been pointed out that the use of a 100 year event is inadequate for urbanised areas (NAE/NRC, 2009). Risk and reliability analyses developed by IPET (2008a) show that a 100 year design for New Orleans is not likely to prevent major ooding for a storm with characteristics comparable to those of Hurricane Katrina. Hurricane return intervals of 4001000 years have been recommended by an independent National Academies review panel as the design basis for protection against storm surges in New Orleans (NRC, 2009). The systemic failure of the New Orleans HPS demonstrates the need for independent peer review during planning and design. Independent, external peer review for large, complex projects has been advocated by National Academies review panels (NRC, 2002; NAE/NRC, 2009). Not only does independent peer review provide fresh thinking about interdisciplinary issues, it also allows for expert opinion about problems that may be politically sensitive for local planning and design teams (NAE/NRC, 2009). It is one of the most effective means of identifying and managing risk, especially for projects involving large, geographically distributed systems. Risk and reliability analyses One of the most signicant accomplishments of the IPET work is the development of a risk and reliability approach for hurricane effects in the New Orleans metropolitan area. Details are provided by IPET (2008a), and only the salient features are described here. In essence, the methodology involves the selection of a suite of scenario hurricanes from a joint frequency distribution, based on historical data, which cover intensity, size, speed, track, landfall location and shape. A suite of 76 scenario hurricanes, each with an annual frequency of occurrence, was chosen to represent the aggregate hurricane hazard in New Orleans, and physics-based computer simulations were performed to determine the storm surge and wave heights associated with each one. The computer simulation process was validated by comparison with measurements and observations from Hurricanes Katrina and Rita. The HPS was divided into multiple reaches, corresponding to lengths of levees and oodwalls with reasonably uniform elevation, soil and groundwater conditions, and hazard, plus features such as pump stations, ood gates and transitions. A fragility curve showing the probability of failure plotted against water elevation was developed for each reach, transition and feature. The fragility curves were dened for the mass stability of the levee or oodwall and foundation soils, as well as erosion after overtopping.

534

OROURKE
10% 1030% 3050% 5070% 7090% 90%

Water volumes entering various areas behind the HPS, referred to as basins and sub-basins, were calculated and linked to the frequency of occurrence of each hurricane. Using event tree analysis, the probabilities of different levels of ooding were estimated and weighted over the 76 scenario hurricanes to assess the annual probabilities of ooding throughout the New Orleans metropolitan area. Flood heights were then related to direct property losses and loss of life. As an example of the results of the risk and reliability analyses, Fig. 42 shows ood-depth frequency maps for the 50 year, 100 year and 500 year hurricanes in the main Orleans East Bank metropolitan basin. The results are shown for the reconstructed HPS as of 2007. Shades of grey display the estimated depths of ooding in 0.6 m increments. The results assume that pumps, which are available to remove water from the basin, are not working, as was the case during Hurricane Katrina. Figure 43 shows the direct property losses resulting from the predicted ooding for a 100 year hurricane for both the pre-Katrina and 2007 HPS, assuming no pumping. The property losses for each sub-basin are shade coded, with black and grey denoting the most severe losses. The 2007 improvements for no pumping have a relatively minor effect on property losses, amounting to about a 7% reduction in losses compared with those associated with the pre-Katrina HPS. The 2007 HPS, however, is shown to provide additional protection when all available pumps are working. Reductions in property losses for the 2007 HPS are approximately 17% lower than those for the pre-Katrina HPS with all pumps working. The reliability and risk analysis shows a substantial reduction in property losses of 35% when all as opposed to no pumps are operating for both the pre-Katrina and 2007 HPS. Interactive maps have been provided by IPET (2008b) at a website that can be accessed by the public to make riskinformed decisions about hurricane ooding potential throughout the New Orleans metropolitan area. Maps are provided for each sub-basin that show ood depths in 0.6 m increments associated with 50 year, 100 year and 500 year hurricanes. A person can view the ood depths in any sub-

(a)

(b)

Fig. 43. Property loss risk maps for 100 year hurricane and no pumping: (a) pre-Katrina; (b) 2007 (IPET, 2008a).

basin for any of the three recurrence intervals, and compare the level of protection for the pre-Katrina, 2007 and 100 year HPS. Lifeline interdependences Figure 44 shows the restoration of electric power customers in the New Orleans and southern Louisiana area as a
Feet of flooding (1 m 328 ft) 8 68 46 24 02

2007, 50 year

2007, 100 year

2007, 500 year

Fig. 42. Examples of ood depth frequency maps for 50 year, 100 year and 500 year hurricanes in New Orleans main basin with no pumping (IPET, 2008a)

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

1 200 000 Katrina landfall 1 000 000 Reconstruction customers Return to service customers Rita landfall 800 000

535

Customers without service

600 000

400 000

200 000

0 LF 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 Restoration days

Fig. 44. Restoration of electric power customers in area affected by Hurricanes Katrina and Rita (after Entergy Services, Inc., 2006)

function of time after both Hurricanes Katrina and Rita. Because of extensive wind damage to electric power lines and ooding of substations, the restoration of electricity proceeded slowly. Twenty-six days after Hurricane Katrina landfall, approximately 20% of customers were not restored, primarily because of property damage and loss of customers who were waiting for reconstruction of their homes or businesses. When Hurricane Rita struck, there was again serious damage and prolonged power outages. After Hurricane Katrina, the supply of crude oil and rened petroleum products was interrupted because of a loss of electric power at the pump stations for three major transmission pipelines: the Colonial, Plantation and Capline pipelines. Fig. 45 shows a map of the Capline, which transports oil from the Louisiana offshore oil port to Illinois for rening and distribution throughout the US Midwest. The Capline is a nominal 1050 mm diameter high-pressure crude oil line, and is one of the largest outside the TransAlaska Pipeline System. As a result of Katrina, major lines of crude oil and rened products were not available for delivery to southern and eastern states, and gasoline and diesel production in the Midwest was seriously affected by lack of supply. About 1.4 million barrels per day of the crude oil supply were lost, accounting for 90% of the production in the Gulf of Mexico, and nearly 160 million litres per day of gasoline production was lost, accounting for approximately 10% of the US supply (ORourke, 2007). The three major pipelines were not fully restored until 17 days after Katrina made landfall. Hurricane Katrina caused the loss of a national liquid fuel delivery system because of electric power failure. This loss is similar to the inability to supply water for reghting because of electric power failure during the 1989 Loma Prieta earthquake, discussed previously. Both these instances illustrate the importance of interdependences among lifeline systems, with electric power being central for inter-operational control of water distribution in San Francisco and liquid fuel transmission in the central US. Especially noteworthy is the scale of fuel supply affected by Hurricane Katrina, which covers thousands of kilometres of pipelines. One of the primary lessons learned from Hurricane Katrina is the vulnerability of critical energy delivery systems as a consequence of hurricanes. At the time of Hurricane Katrina, Gulf of Mexico infrastructure provided over 60% of port facilities for US oil imports, nearly 47% of US rening

Transmission pipelines Patoka terminal

Capline

Port Fourchon Louisiana offshore oil port 0

N 500 km

Fig. 45. Map of Capline and associated liquid fuel transmission pipelines

capacity, and approximately 19% and 29% of US natural gas and oil production respectively (US Energy Information Administration, 2005). MISSISSIPPI RIVER AND GULF OF MEXICO GEOSYSTEMS Figure 46 shows a cross-section through New Orleans near its central business district. The cross-section is shown at exaggerated scale to emphasise the basin characteristics of the city. The approximate pre-Katrina ood design water

536
Lake Pontchartrain Gentilly Ridge A'

OROURKE

Lake Borgne

(a) Hurricane protection levee and floodwall 53 m Hurricane surge level 35 m Drainage area of Mississippi and Missouri Rivers N New Orleans Mississippi delta 0 km 500

Floodwall along 9 m Mississippi River 6m 3m 7m 55 m Mississippi flood level

Fig. 47. Mississippi River drainage system

ive r

0m 3 m 6 m

Gentilly Ridge

Mi ss

Interstate 10 St Louis Cathedral (b)

Dillard University

Fig. 46. New Orleans cross-section: (a) aerial view; (b) crosssection AA9

levels in metres according to the National Geodetic Vertical Datum are shown for the Mississippi River and Lake Pontchartrain. Parts of the city are as much as 56 m below hurricane surge and river ood levels. Moreover, subsidence in the New Orleans metropolitan area varies from 6 to 17 mm/year, with about 10 mm/year in Orleans Parish near Lake Pontchartrain (IPET, 2008a), further compounding the differential between high water and the basin ground surface. New Orleans is located between a proverbial rock and a hard place, with the rock being the river and the hard place being the sea. It is therefore instructive to examine more closely the Mississippi River and Gulf of Mexico geosystems, which set the ultimate boundary conditions for New Orleans and the entire Mississippi delta. Mississippi River and delta The Mississippi River system (including the Missouri River and its tributaries) drains 41% of the US and parts of Canada. As illustrated in Fig. 47, the total drainage area covers 3 227 000 km2 (Kaufman, 1978). The ofcial peak ow measured by the USACE in 1927 was 72 000 m3 /s, although maximum ow measured by others is reported to be as high as 85 000 m3 /s (Barry, 1998). To put this in perspective, peak ow of the Mississippi River in one minute exceeds the maximum daily New York City water supply of 4 billion litres. The river carries not only water but also sediment, crucial for sustaining the delta and maintaining the wetlands that act as a buffer against hurricane surge. The sediment load in the Mississippi River just north of the Atchafalaya River is approximately 228 million t/year (Keown et al., 1981). This volume of sediment is enough to cover 150 km2 to a depth of about 1 m. The Mississippi River delta covers an area of about 25 000 km2 (Day et al., 2007), referred to as the Mississippi deltaic plain (MDP). The MDP is in dynamic equilibrium with the Gulf of Mexico through the complex interaction of sedimentation, vegetation, erosion, and human development.

The MDP formed about 60007000 years ago, and consists of a series overlapping deltaic lobes that were deposited in response to changes in the main river channel location. Fig. 48 presents a satellite image of the MDP developed by Day et al. (2007). Each lobe is numbered in the gure caption with its name and approximate time of formation before present (BP). The gure shows that the Mississippi River has changed course about every 700 to 1800 years. The river course alterations are the natural means of restoring and maintaining the MDP. They are punctuated, however, by catastrophic changes in main channel location, which may involve distances of several hundred km. Figure 49 is a map of southeastern Louisiana, showing the Atchafalaya and Mississippi Rivers. The Atchafalaya River is the fth largest river in the US, as measured by discharge (Reuss, 2004). Its average gradient is over two times steeper than that of the Mississippi, and if unchecked would erode into the main channel of the Mississippi River and provide a hydraulically more efcient path to the sea. Between 1900 and 1950 the amount of the Mississippi ow into the Atchafalaya increased from 12% to 30% (Reuss, 2004). By 1954 the Atchafalaya River had captured about one third of the ow of the Mississippi River. At that time the US Congress authorised the USACE to build a protective structure to regulate ow and sediment transfer between the two rivers. In 1963 the Old River control structure was commissioned by the USACE. Its approximate location at the junction

a ke ch La ont P

iss

ipp

iR

ai r tr n

si Mis

ss

N
ipp

i River

0 New Orleans

25 km

1 Atchafalaya Bay

5 2 4 6

Gulf of Mexico

Mississippi bird foot delta

Fig. 48. Mississippi River deltaic plain showing overlapping depositional lobes caused by different river channels. Times of deposition: (1) 4600 years BP; (2) 35002800 years BP; (3) 28001000 years BP; (4) 1000300 years BP; (5) 750500 years BP; (6) 550 years BP to current times (after Day et al., 2007)

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Louisiana Old River control structure Mississippi

537

Baton Rouge

New Orleans

20 m

Gulf of Mexico 0 100 m

N 50 km

Fig. 49. Mississippi River, Atchafalaya River and Mississippi delta

Red River LOUISIANA

Mississippi River MISSISSIPPI Sydney A. Murray power plant Overbank structure

In the ood of 1973 the Old River structure nearly failed. Severe erosion around the low sill structure destroyed its southern wing wall and scoured the soil surrounding the pile-supported main portion of the facility. There is surprisingly little written in the technical literature about the performance of this facility, although it is graphically described in journalistic accounts (e.g. McPhee, 1989). Water scoured a 16 m deep hole on the Mississippi side of the low sill structure, and threatened to enlarge a hole approximating the size of a football stadium that had developed on the Atchafalaya side during the previous decade (Reuss, 2004). The foundation system for the lower sill structure was restored by dumping riprap into the scour zones, grouting, and changing the maximum difference in head from 11.3 to 6.7 m between the Mississippi and Atchafalaya sides of the sill. An additional facility, known as the auxiliary structure, was constructed and commissioned in 1986 to supplement the control system. This structure is shown in Fig. 50 just south of the low sill structure. Failure of the Old River structure in 1973 would have had serious economic and social consequences. Some of these consequences have been summarised by Kazmann & Johnson (1980): they include catastrophic ooding of the Atchafalaya River basin, failures of bridges and associated effects on regional highway and rail transportation, failure of pipelines and associated effects on energy transmission systems, and loss of fresh water supply for drinking and industrial purposes along the lower Mississippi River. The most important effect is loss of the main river course to New Orleans and Baton Rouge and the attendant impact on the commerce, industry and social fabric of southeastern Louisiana. Gulf of Mexico and Louisiana coast protection Historically, protection against Mississippi oods and hurricane surge has been a vicious circle. The approach to controlling the river has been to build increasingly larger levees, and close outlets to distributaries. This in turn has restricted the distribution of sediment, leaving wetlands and barrier islands without soil to counteract storm and hurricane erosion. Loss of wetlands and barrier islands has increased exposure to storm surge and hurricane ooding. From 1875 to 1927 Mississippi River ood protection was dominated by a policy of constructing higher levees and closing ow to distributaries (Barry, 1998). By conning ow and sediment transport to the main channel of the river, the natural mechanisms for delta restoration by ow through distributaries and overbank ooding were blocked, with the resulting loss of most sediment to the deep waters of the Gulf of Mexico. In the lower Mississippi Valley alone there are 2560 km of levees along the Mississippi River (Kaufman, 1978). Sediment transport is still conned predominantly to the main river channel. Moreover, the construction of several large dams on the Missouri River and the construction of the Missouri River Bank Stabilization and Navigation Project have greatly reduced sediment deliveries to the Mississippi (NRC, 2009). From 1964 to 1981 the measured sediment load in the lower Mississippi River decreased by nearly 50%, from 388 million to 228 million t/ year (Keown et al., 1981). On the Gulf side, New Orleans and the Mississippi delta are at increasing risk from storms and hurricanes. In 2005 there were 28 major named storms in the North Atlantic Ocean (NAO), the largest number on record. Fourteen of these storms were hurricanes, of which seven were major hurricanes with maximum sustained wind speed of at least 178 km/h (e.g. Rauber et al., 2005). The threat of storms is compounded by a rising sea level. The Intergovernmental

r ive R pi ip ss si is r M ive aR lay afa ch At

N 0 km 5 Old River lock Atchafalaya River

Low sill structure Overflow channel Auxiliary structure Old River

LOUISIANA

Fig. 50. Old River structure

between the Mississippi and Atchafalaya Rivers is shown in Fig. 49, and an expanded view of it is shown in Fig. 50. As indicated in the gure, the Red River joins the Atchafalaya at the Old River, which is a remnant meander of the Mississippi. When originally constructed, the Old River structure consisted of a low sill structure, which acts as a weir to control ow into the Atchafalaya at any stage of the Mississippi River. An overbank structure was also constructed to divert water during ood stages of the Mississippi. Water is conveyed to the Atchafalaya through an overow channel. The project also involved construction of a navigation lock, which is shown in the gure as the Old River lock.

538

OROURKE for risk-based loading assessments. It involves historical trends, institutional cultures, community values, and interdisciplinary communication that helps set acceptable levels of risk and guide the structural and non-structural measures for risk reduction. Hurricane Katrina also reveals issues at an even larger scale. The sustainability of New Orleans and the Mississippi delta depends on the control of oods and management of sediment in a 3 227 000 km2 drainage area, the prevention of catastrophic change in the course of the Mississippi River through capture by the Atchafalaya River, the distribution of sediment through a 2500 km2 deltaic plain to maintain wetlands and barrier islands, the construction and management of hurricane protection facilities that are coordinated with effective coastal restoration, and preparations that account for sea level rise and the potential for increased storm frequency and intensity in the Gulf of Mexico. As previously discussed, the solutions to these overarching problems are frequently in conict with each other. To resolve these conicts engineers need to focus increasingly on trade-offs that address the physical, environmental and human uncertainties that are inherently part of geosystems and infrastructure. Planning for geosystems needs to recognise the limitations of sustainability en masse, and make choices about allocating limited resources, restricting highrisk developments, and moving communities from vulnerable locations. Selective sustainability may be the best policy, in which high-priority projects are implemented within the existing budget of resources to achieve well-dened economic and societal benets.

Panel on Climate Change (IPCC, 2008) reports a global average sea level rise of 1.8 mm/year from 1961 to 2003, and an average rise of 3.8 mm/year from 1993 to 2003. Moreover, subsidence in the Mississippi delta, as described previously for the New Orleans area, increases the differential settlement relative to sea surface. The IPCC also reports an increase in intense tropical cyclone activity in the NAO since about 1970. Holland & Webster (2007) show that the increase in NAO cyclonic activity correlates with an increase in the sea surface temperature (SST), with a 100% increase in tropical cyclone and hurricane numbers corresponding to an increase of 0.78C in SST. Loss of sediment from the Mississippi River exacerbates the problems associated with erosion and loss of wetlands related to storms and increased sea level in the Gulf of Mexico. From 1990 to 2000 wetland loss in the MDP was approximately 62 km2 /year, and is forecast over the next 50 years under current restoration policy to be approximately 1300 km2 (Barras et al., 2003; NRC, 2009). Recognising the integrated effects of ood control, sediment transport and hurricane protection, the USACE after Hurricane Katrina was authorised through public law to develop and present a full range of ood control, coastal restoration, and hurricane protection measures to provide . . . protection for a storm surge equivalent to a Category 5 hurricane (NRC, 2009). In response to the federal mandate, the Corps issued the Louisiana Coastal Protection and Restoration (LACPR) Final Technical Report (USACE, 2009), which was developed in coordination with the State of Louisiana master plan for coastal protection and restoration (CPRA, 2007). The LACPR study developed 111 alternative risk reduction plans for ve areas of the Louisiana coast and delta system. An independent National Academies review panel (NRC, 2009) has raised concerns about the LACPR plan. Among the institutional issues of concern are the lack of a unied plan and priority projects, the piecemeal approach to project authorisation, the lack of a convincing plan to prevent development in high-risk areas, the need for more rigorous multi-criteria decision-making, and the absence of a comprehensive systems approach. Among the technical issues of concern are the need to quantify uncertainties associated with wetland restoration and river diversions, a lack of clarity about the impact of river diversions on navigation, insufcient treatment of potential structural failures, and questions about the availability of adequate sediment to revive coastal systems. A key assumption of the LACPR study is that the current conguration of the Louisiana coastline is sustainable (USACE, 2009). The National Academies review panel raises concerns about the technical feasibility of this assumption, especially given the reductions in Mississippi sediment load and the increasing loss of wetlands from storms and rising sea level. Rather than consuming nancial resources on maintaining the current coastal system en masse, the NRC committee encourages the LACPR team to focus its plans on high-priority projects. Summary A review of Hurricane Katrinas effects on the New Orleans HPS and the Gulf of Mexico oil and natural gas production and transportation infrastructure reveals problems of immense scale and complexity. The 565 km long system of levees and oodwalls, for example, involves geotechnical problems of subsidence, erosion, site characterisation, and HPS loading and failure mechanisms under diverse and difcult soil conditions. It involves storm surge and wave predictions and their potential frequency at various locations

CONCLUSIONS To address the issue of scale in geotechnical engineering for geohazards and large, geographically distributed systems, a generalised model is proposed in which there are six divisions of scale, ranging from nano scale (109 m) at the molecular level to mega scale (106 m) at the level of geographic regions. The paper explores nine orders of magnitude of scale with respect to the modelling of large, geographically distributed systems, from the behaviour of partially saturated sand to the earthquake performance of the water supplies in San Francisco and Los Angeles, and the hurricane performance of the ood protection system in New Orleans. A signicant trend in geotechnical engineering for component behaviour has been the implementation of large-scale testing facilities for soilstructure interaction, and the following conclusions can be drawn from the results of largescale split-box testing of pipeline response to earthquakeinduced ground rupture, (a) For the test sands described in this paper, the size of the DS apparatus has a signicant effect on the determination of the DS peak friction angle, d 9s-p . Values of d 9s-p for the 60 mm DS apparatus were 468 higher than those for the 100 mm DS apparatus, and 6 78 higher than those for the 300 mm DS apparatus. The difference in d 9s-p is related to the increased angle of dilatancy, p , for the 60 mm device. The 100 mm device, fabricated according to the design of Lings & Dietz (2004), gives the most reliable results, and is recommended for DS testing. (b) A procedure is presented for using DS test measurements to determine the plane-strain friction angle at peak DS strength, ps-p , and the plane-strain cohesive intercept cps . The cohesion is directly related to the increased suction-induced dilatancy of partially saturated sand relative to that of dry sand.

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS (c) There is excellent agreement between FE simulations of the horizontal force against displacement of buried pipe using ps-p and cps determined from DS tests on partially saturated sand and the analytical results of full-scale 2D tests for a wide range of pipe depth to diameter ratios. These results provide supporting evidence for the reduction of ps-p and cps from DS data as presented in this paper. (d ) The large-scale test data conrm the substantial ductility of HDPE pipe and its benecial effects in accommodating permanent ground deformation. The maximum measured strains of 8% for 1.2 m of strikeslip displacement were far below strain levels associated with pipe wall rupture. The maximum reduction of pipe diameter due to ovalling, however, was 12%, indicating that loss of pipe cross-sectional area is likely to govern failure of HDPE pipes with large D/t for ground rupture effects. The evaluation of water supply response to earthquakes in San Francisco and Los Angeles, and of the ood protection system response to hurricanes in New Orleans, provides many lessons with respect to complex system performance, and the following conclusions can be drawn with respect to systems modelling and infrastructure protection: (e) Computer simulations of water supply performance during the 1906 San Francisco earthquake underscore the critical importance of liquefaction hazards, and the interactions among liquefaction-induced ground deformation, water distribution system performance, and catastrophic re and destruction of the built environment. Simulations of the water supply response to earthquake effects before the 1989 Loma Prieta earthquake were instrumental in developing alternative water supply sources and emergency procedures that were actually used to suppress re and prevent conagration, even though water ow in underground pipelines had been cut off by liquefaction-induced ground deformation. ( f ) The systematic analysis of pipeline repair records after the 1994 Northridge earthquake shows the locations of important seismic and geotechnical hazards. They are used to identify zones of potential ground failure not recognised in previous explorations, and to develop regressions linking damage rates and various levels of strong motion that are important for earthquake loss estimation studies. (g) A hydraulic network model has been developed for earthquake effects on the Los Angeles water supply that accounts for all 11 700 km of its water trunk and distribution pipelines and related facilities, an ensemble of 59 scenario earthquakes, and the effects of transient and permanent ground deformation on underground pipelines. Simulations performed for the Los Angeles water supply response to a 7.8 MW earthquake on the southern San Andreas fault show that opening reservoirs that have been disconnected because of water quality regulations improves serviceability signicantly in the locations of highest population, and is thus an effective means of emergency response, even though such action will require tap water safety notices to be issued for the entire system. (h) The effects of Hurricane Katrina on New Orleans and the Mississippi delta demonstrate the problems resulting from inadequate characterisation of system components and the absence of an integrated assessment of geohazards. Substantial and important work has been accomplished after Hurricane Katrina in evaluating the

539

(i)

HPS and taking corrective measures to develop resilient ood protection. The HPS performance in response to Hurricane Katrina and its reconstruction represents an indispensable case history with respect to geohazards and large, geographically distributed systems. One of the most signicant accomplishments of the postKatrina work is the development of a comprehensive risk and reliability approach for hurricane effects, which is described in this paper. Flooding of the Mississippi, hurricanes in the Gulf of Mexico, and the sustainability of the Mississippi deltaic plain and Louisiana coastline involve engineering at an unprecedented scale, not only with respect to sheer geographic size, but also with regard to the complexities of the physical processes, environmental systems, and human communities embedded in the river, delta and coastal areas. Harmonising an approach and rational plan under these conditions is a work in progress. There are several unresolved issues that are part of this process, including lack of consensus on the mechanism for soil failure at the 17th Street canal, selection of appropriate safety factor and hurricane recurrence interval for HPS design in urbanised areas, and an integrated ood prevention plan for the Mississippi River and Gulf of Mexico that involves realistic assumptions about sediment distribution, coastal restoration, and protection against catastrophic changes in the course of the Mississippi River.

Analysing the response of infrastructure to geohazards teaches us about the interdependences among critical infrastructure systems, their inherent vulnerabilities, and improved methods for modelling and managing them. Evaluating the effects of geohazards on geographically distributed systems teaches us about the characterisation of natural hazards, modelling of their spatial and temporal variability, and methods for making risk-based decisions under conditions of high uncertainty. The lessons from this reciprocal learning process are important for the life safety and economic viability of communities worldwide. In the future, large, geographically distributed systems will be planned and managed with increasingly more complex, risk-based models. Geotechnical engineering plays an important and indispensable role in this modelling and management activity, which is inherently multidisciplinary. The degree to which the geotechnical community learns from geohazards and develops the skills necessary for solving problems on the scale of geosystems will determine the level of leadership it will exercise in addressing some of the most important challenges of the next century. ACKNOWLEDGEMENTS I gratefully acknowledge Dr David J. Henkel, the 22nd Rankine Lecturer, and Dr Melvin I. Esrig for their example and encouragement as teachers at Cornell University and their lifelong friendship. I also thank Drs Edward J. Cording and Ralph B. Peck, former professors at the University of Illinois, for their teaching, exemplary professional conduct, and encouragement. Part of the work, on which this paper is based, was supported by the George E. Brown, Jr Network for Earthquake Engineering Simulation (NEES) Program of the National Science Foundation (NSF) under Grant No. CMS0421142 and the Multidisciplinary Center for Earthquake Engineering Research (MCEER) and NSF under award EEC-9709471. Any opinions, ndings and conclusions or recommendations expressed in this material are those of the author and do not necessarily reect the views of the NSF.

540

OROURKE
3 9 p p 9 ps p 9s y 9x , x 9y crit c 9rit dsp d 9sp p 9s psp p minor principal total stress total shear stress effective shear stress peak total shear stress peak effective shear stress plane-strain total shear stress plane-strain effective shear stress effective shear stress parallel and normal to the DS rupture plane, respectively total stress friction angle at constant volume during shear deformation critical-state effective stress friction angle total stress DS peak friction angle effective stress DS peak friction angle effective stress plain-strain friction angle total stress plane-strain friction angle at peak DS strength angle of dilatancy peak angle of dilatancy

Thanks are extended to the Cornell NEES equipment site team, including Professor Harry Stewart, Dr Mike Palmer, and Messrs Tim Bond and Joe Chipalowsky. I thank in particular Dr Nate Olson and Jai Jung for their help and contributions to this paper. The support of the Los Angeles Department of Water and Power, including Dr Craig Davis, Mr Glenn Singley, Dr Jianping Hu, and Mr Victor Vargas, is gratefully acknowledged. Frank Blackburn, former Deputy Fire Chief of the San Francisco Fire Department, and Dr Charles Scawthorn provided advice regarding re risk in San Francisco. Professor Kenichi Soga and Dilan Robert of the University of Cambridge provided help in modelling the strain-softening of soil for FE analyses of soilpipeline interaction. Dr Ed Link helped in obtaining information about the New Orleans HPS and details of the IPET study used in this paper.

NOTATION
Cu c cds cps D Dmax D50 d FE dxy, dx p, dxf d d =2 E F F9 H Hc H bk L n Nq r 2 , R2 Suvst t um w Y Y9 tan yy d , dry p f t yx , xy x y 9 N N 9 coefcient of uniformity intercept of linear regression of tops versus centres of Mohr circles from DS tests; cohesive intercept total stress DS cohesive intercept total stress plane-strain cohesive intercept pipe external diameter maximum particle size mean grain size FE element size DS test horizontal displacements at yield, peak strength, and crit at which full softening occurs normal strain increment shear strain increment Youngs modulus horizontal force dimensionless horizontal force box height in DS apparatus; thickness of DS specimen pipe centreline depth distance between pipe and closest test basin wall horizontal dimension of DS testing device; pipe length number of test specimens dimensionless lateral force coefcient of determination undrained vane shear strength pipe wall thickness matric suction gravimetric water content lateral displacement of pipe with respect to soil dimensionless lateral pipe displacement slope of linear regression of tops versus centres of Mohr circles from DS tests strain strain normal to DS rupture plane shear strain dry soil unit weight shear strain increment beyond yield at which there is no dilation total soil unit weight shear strain parallel and normal to the DS rupture plane, respectively horizontal DS soil displacement vertical DS soil displacement total stress effective stress total normal stress effective normal stress plane-strain total normal stress plane-strain effective normal stress vertical total stress at pipe centre effective stress in direction of DS rupture plane effective stress normal to DS rupture plane major principal total stress

REFERENCES
ABAQUS (2006). Users manual. Providence, RI: ABAQUS, Inc. American Society for Testing and Materials (ASTM) (2008a). D6938-07b: Standard test method for in-place density and water content of soil and soil-aggregate by nuclear methods (shallow depth). Annual Book of ASTM Standards Vol. 4.09. West Conshohocken, PA: ASTM International. American Society for Testing and Materials (ASTM) (2008b). D2216-98: Standard test methods for laboratory determination of water (moisture) content of soil and rock by mass. Annual Book of ASTM Standards, Vol. 4.08. West Conshohocken, PA: ASTM International. American Society for Testing and Materials (ASTM) (2008c). D6836-02: Standard test methods for determination of the soil water characteristic curve for desorption using a hanging column, pressure extractor, chilled mirror hygrometer, and/or centrifuge. Annual Book of ASTM Standards, Vol. 4.09. West Conshohocken, PA: ASTM International. American Society for Testing and Materials (ASTM) (2008d). D3080-04: Standard test methods for direct shear test of soils under consolidated drained conditions. Annual Book of ASTM Standards, Vol. 4.08. West Conshohocken, PA: ASTM International. American Society of Civil Engineers (ASCE) (1984). Guidelines for the seismic design of oil and gas pipeline systems. Reston, VA: ASCE, Committee on Gas and Liquid Fuel Lifelines. Anastasopoulos, G., Gazetas, M. F., Bransby, M. C. R., Davies, A. & El Nahas, A. (2007). Fault rupture propagation through sand: nite-element analysis and validation through centrifuge experiments. J. Geotech. Geoenviron. Engng ASCE 133, No. 8, 943 958. Armando, L. (1987). Handbook of hydraulic engineering. New York, NY: John Wiley & Sons. ASCE Hurricane Katrina External Review Panel (2007). The New Orleans hurricane protection system: What went wrong and why. Reston, VA: ASCE. Barras, J., Beville, S., Britsch, D., Hartley, S., Hawes, S., Johnson, J., Kemp, P., Kinler, Q., Martucci, A., Porthouse, J., Reed, D., Roy, K., Sapkota, S. & Suhayda, J. (2003). Historic and predicted coastal Louisiana land changes: 19782050, US Geological Survey Open File Report 03-334. Baton Rouge, LA: US Geological Survey, National Wetlands Research Center. Barry, J. M. (1998). Rising tide. New York, NY: Touchstone Books, Simon & Schuster. oBolton, M. D. (1986). The strength and dilatancy of sands. Ge technique 36, No. 1, 6578, doi: 10.1680/geot.1986.36.1.65. Bonneau, A. & ORourke, T. D. (2009). Water supply performance during earthquakes and extreme events, MCEER-03-0009. Buffalo, NY: MCEER, University of New York at Buffalo. Brandon, T. L., Wright, S. G. & Duncan, J. M. (2008). Analysis of the stability of I-walls with gaps between the I-wall and the levee ll. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 692700. Bray, J. D., Seed, R. B., Cluff, L. S. & Seed, H. B. (1994).

ps , Nps p 9s , p 9N vc 9 xx 9 yy 1

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Earthquake fault rupture propagation through soil. J. Geotech. Engng ASCE 120, No. 3, 543561. British Standards Institution (BSI) (1990). BS 1377: Part 7: 4: Determination of shear strength by direct shear. Milton Keynes: British Standards Institution. Burby, R. (2001). Flood insurance and oodplain management: the US experience. Global Environmental Change Part B: Environmental Hazards 3, Nos 34, 111122. Cerato, A. B. & Lutenegger, A. J. (2006). Specimen size and scale effects of direct shear box tests on sands. Geotech. Test. J. 29, No. 6, 507516. Chong, K. P. & Davis, D. C. (1999). Engineering mechanics and materials research in the information technology age. Appl. Math. Mech. 20, No. 12, 13051309. Christian, J. T. (2007). Lessons from Hurricane Katrina. The Bridge, Spring, 511. Christian, J. T. (ed.) (2008). Performance of geo-systems during Hurricane Katrina (Special issue). J. Geotech. Geoenviron. Engng ASCE 134, No. 5. Coastal Protection and Restoration Authority of Louisiana (CPRA) (2007). Louisianas comprehensive plan for a sustainable coast. Coastal Protection and Restoration Authority of Louisiana, State of Louisiana. http://www.lacpra.org/ Cosgrave, J. (2007). Synthesis report: Expanded summary. Joint evaluation of the international response to the Indian Ocean tsunamis. London: Tsumani Evaluation Coalition. Dane, J. H. & Topp, G. C. (2002). Methods of soil analysis. Part 4: Physical methods. Madison, WI: Soil Science Society of America. Davis, E. H. (1968). Theories of plasticity and the failure of soil masses. In Soil mechanics: selected topics (ed. I. K. Lee), pp. 341380. London: Butterworth. Day, J. W. Jr, Boesch, D. F., Clairain, E. J., Kemp, G. P., Laska, S. B., Mitsch, W. J., Orth, K., Mashriqui, H., Reed, D. J., Shabman, L., Simenstad, C. A., Streever, B. J., Twilley, R. R., Watson, C. C., Wells, J. T. & Whigham, D. F. (2007). Restoration of the Mississippi delta: lessons from Hurricanes Katrina and Rita. Science 315, No. 5819, 16791684. Dietz, M. S. (2000). Developing an holistic understanding of interface friction using sand within the direct shear apparatus. PhD thesis, University of Bristol, UK. Duncan, J. M., Brandon, T. L., Wright, S. G. & Vroman, N. (2008). Stability of I-walls in New Orleans during Hurricane Katrina. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 668691. Dunbar, J. B. & Britsch, L. D. III (2008). Geology of the New Orleans area and the canal levee failures. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 566582. Earthquake Engineering Research Institute (EERI) (2008). The Wenchuan, Sichuan Province, China earthquake of May 12, 2008, EERI Special Earthquake Report, October 2008. Oakland, CA: EERI. Eguchi, R. T., Goltz, J. D., Taylor, C. E., Chang, S. E., Flores, P. J., Johnson, L. A., Seligson, H. A. & Blais, N. C. (1996). The Northridge earthquake as an economic event: direct capital losses, analyzing economic impacts and recovery from urban earthquake: issue for policy makers. Proceedings of the EERI Conference, Pasadena, CA, pp. 128. Ellsworth, W. L. (1990). Earthquake history, 17691989. In The San Andreas Fault System, USGS Professional Paper 1515, pp. 153188. Washington, DC: US Government Printing Ofce. Emanuel, K. (2005). Divine wind. New York, NY: Oxford University Press. Environmental Protection Agency (EPA) (2008). EPANET. Washington, DC: Environmental Protection Agency (http://www.epa.gov/nrmrl/wswrd/dw/epanet.html). Entergy Services, Inc. (2006). Electric power customer restoration plots provided by Utility Operations Communications Department. Little Rock, AR: Energy Services, Inc. Federal Emergency Management Agency (FEMA) (2006). Multi hazard loss estimation methodology earthquake model, HAZUSMH MR3. Technical Manual. Washington, DC: Department of Homeland Security, FEMA Mitigation Division (http://www.fema.gov/plan/prevent/hazus/hz_manuals.shtm). Fitzpatrick, P. J. (2006). Hurricanes. Santa Barbara, CA: ABCCLIO, Inc. Fradkin, P. L. (2005). The great earthquake and restorms of 1906. Berkeley, CA: University of California Press.

541

Hamada, M., Isoyama, R. & Wakamatsu, K. (1995). The 1995 Hyogoken-Nambu (Kobe) earthquake liquefaction, ground displacement and soil condition in Hanshin area. Tokyo: Association for Development of Earthquake Prediction, Waseda University. Hoeksema, R. (2006). Designed for dry feet ood protection and land reclamation in the Netherlands. Reston, VA: ASCE. Holland, G. J. & Webster, P. J. (2007). Heightened tropical cyclone activity in the North Atlantic: Natural variability or climate trend? Phil. Trans. R. Soc. London A 365, 26952716. Holzer, T. L., Bennet, M. J., Ponti, D. J. & Tinsley, J. C. III (1999). Liquefaction and soil failure during 1994 Northridge earthquake. J. Geotech. Geoenviron. Engng ASCE 125, No. 6, 438452. Independent Levee Investigation Team (2006). Investigation of the New Orleans regional ood protection systems during hurricane Katrina: Final report (http://www.ce.berkeley.edu/,new_ orleans/). Interagency Performance Evaluation Task Force (IPET) (2008a). Performance evaluation of the New Orleans and southeast Louisiana hurricane protection system: Final report. US Army Corps of Engineers (https://ipet.wes.army.mil/). Interagency Performance Evaluation Task Force (IPET) (2008b). Risk and reliability report. US Army Corps of Engineers (http:// nolarisk.usace.army.mil/) Intergovernmental Panel on Climate Change (IPCC) (2008). Climate change 2007: Synthesis report. Geneva: Intergovernmental Panel on Climate Change (http://www.ipcc.ch/). Iwan, W. D. (ed.) (2006). The great Sumatra earthquakes and Indian Ocean tsunamis of 26 December 2004 and 28 March 2005. Earthquake Spectra Special Issue III, 22, 112. Jeon, S.-S. & ORourke, T. D. (2005). Northridge earthquake effects on pipelines and residential buildings. Bull. Seismol. Soc. Am. 95, No. 1, 125. Jewell, R. A. & Wroth, C. P. (1987). Direct shear tests on otechnique 37, No. 1, 5368. reinforced sand. Ge Johnston, A. C. & Schweig, E. S. (1996). The enigma of the New Madrid earthquakes of 18111812. Ann. Rev. Earth Planet. Sci. 24, 339384. Jordan, J. W. & Paulius, S. L. (2006). Lessons learned from Hurricane Katrina. Proc. Forensic Engng 2006, Reston, VA, 4169. Kafali, C. & Grigoriu, M. (2007). Seismic fragility analysis: application to simple linear and nonlinear systems. Earthquake Engng Struct. Dynam. 36, No. 13, 18851900. Kaufman, R. I. (1978). Control of 1973 Mississippi River ood. J. Water Res. Plan. Mngmt Div. ASCE 104, No. WR1, 105121. Kazmann, G. & Johnson, D. B. (1980). If the old river structure fails, Bulletin 12. Baton Rouge, LA: Louisiana Water Resources Research Institute. Keown, M. P., Dardeau, E. A. & Causey, E. M. (1981). Characterization of the suspended-sediment regime and bed-material gradation of the Mississippi River basin: Final report. Vicksburg, MS: US Army Engineer Waterways Experiment Station. Khater, M. M., Grigoriu, M. D. & ORourke, T. D. (1989). Serviceability measures and sensitivity factors for estimating seismic performance of water supply systems. Proc. 9th World Conf. Earthquake Engng, Tokyo 8, 123128. Lawson, A. C., Gilbert, G. K., Reid, H. F., Branner, J. C., Fairbanks, H. W., Wood, H. O., Hayford, J. F., Baldwin, A. L., Omori, F., Leuschner, A. O., Davidson, G., Matthes, F. E., Anderson, R., Louderback, G. D., Holway, R. S., Eakle, A. S., Crandall, R., Hoffman, G. F., Warring, G. A., Hughes, E., Rogers, F. J. & Baird, A. (1908). The California earthquake of April 18, 1906: report of the California State Earthquake Investigation Commission, Publication No. 87 (two volumes and atlas). Washington, DC: Carnegie Institute. Lings, M. L. & Dietz, M. S. (2004). An improved direct shear otechnique 54, No. 4, 245256, doi: apparatus for sand. Ge 10.1680/geot.2004.54.4.245. Longshore, D. (1998). Encyclopedia of hurricanes, typhoons, and cyclones. New York, NY: Facts On File, Inc. Manson, M. (1908). Reports on auxiliary water supply system for re protection for San Francisco, California. San Francisco, CA: Board of Public Works. McPhee, J. (1989). The control of nature. New York, NY: Farrar, Straus, and Giroux. Minerals Management Service (MMS) (2006). Impact assessment of offshore facilities from Hurricanes Katrina and Rita, Release

542

OROURKE
the Atchafalaya Basin 18001995. College Station, TX: Texas A&M University Press. Rogers, J. D., Boutwell, G. P., Schmitz, D. W., Karadeniz, D., Watkins, C. M., Athanasopoulos-Zekkos, A. G. & Cobas-Boa, D. (2008). Geologic conditions underlying the 2005 17th Street canal levee failure in New Orleans. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 583601. Romero, N., ORourke, T. D., Nozick, L. K. & Davis, C. R. (2010). Seismic hazards and water supply performance. J. Earthquake Engng (accepted for publication). Sasanakul, I., Vanadit-Ellis, W., Sharp, M., Abdoun, T., Ubilla, J., Steedman, S. & Stone, K. (2008). New Orleans levee system performance during Katrina: 17th Street Canal and Orleans Canal North. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 657667. Scawthorn, C., ORourke, T. D. & Blackburn, F. B. (2006a). The San Francisco earthquake and re of 1906: enduring lessons for re protection and water safety. Earthquake Spectra 22, No. 52, 51355158. Scawthorn, C., Korneld, L., Seligson, H. & Rojahn, C. (2006b). Estimated losses from scenario earthquakes affecting San Francisco: CAPPS: Part 2. Proc. 8th US Conf. Earthquake Engng, San Francisco, CA, paper no. 1595 (CD-ROM). Schoeld, A. (2005). Disturbed soil properties and geotechnical design. London: Thomas Telford. Schussler, H. (1906). The water supply of San Francisco, California. New York, NY: Martin B. Brown Press. Seed, R. B., Bea, R. G., Abdelmalak, R. I., Athanasopoulos-Zekkos, A., Boutwell, G. P., Briaud, J.-L., Cheung, C., Cobos-Roa, D., Ehrensing, L., Govindasamy, A. V., Harder, L. F. Jr, Inkabi, K. S., Nicks, J., Pestana, J. M., Porter, J., Rhee, K., Riemer, M. F., Rogers, J. D., Storesund, R., Vera-Grunauer, X. & Wartman, J. (2008a). New Orleans and Hurricane Katrina I: Introduction, overview and the east ank. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 701717. Seed, R. B., Bea, R. G., Athanasopoulos-Zekkos, A., Boutwell, G. P., Bray, J. D., Cheung, C., Cobos-Roa, D., Ehrensing, L., Harder, L. F. Jr, Pestana, J. M., Riemer, M. F., Rogers, J. D., Storesund, R., Vera-Grunauer, X. & Wartman, J. (2008b). New Orleans and Hurricane Katrina II: The Central Region and the Lower Ninth Ward. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 718739. Seed, R. B., Bea, R. G., Athanasopoulos-Zekkos, A., Boutwell, G. P., Bray, J. D., Cheung, C., Cobos-Roa, D., Harder, L. F. Jr, Moss, R. E. S., Pestana, J. M., Riemer, M. F., Rogers, J. D., Storesund, R., Vera-Grunauer, X. & Wartman, J. (2008c). New Orleans and Hurricane Katrina III: The 17th Street Drainage Canal. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 740761. Shibuya, S., Mitachi, T. & Tamate, S. (1997). Interpretation of otechdirect shear box testing of sands as quasi-simple shear. Ge nique 47, No. 4, 769790, doi: 10.1680/geot.1997.47.4.769. Shinozuka, M., Feng, M. Q., Lee, J. & Naganuma, T. (2000). Statistical analysis of fragility curves. J. Engng Mech. ASCE 126, No. 12, 12241231. Stone, R. (2009). A deeply scarred land. Science 324, 8 May, 713 714. Tokimatsu, K. & Suzuki, H. (2004). Pore water pressure response around pile and its effects on p y behavior during soil liquefaction. Soils Found. 44, No. 6, 101110. Toprak, S., ORourke, T. D. & Tutnucu, I. (1999). Geographic information system (GIS) characterization of spatially-distributed lifeline damage. Proc. 5th US Conf. on Lifeline Earthquake Engng, Washington, DC, 110119. Trautmann, C. H. (1983). Behavior of pipe in dry sand under lateral and uplift loading. PhD thesis, Cornell University, Ithaca, NY. Trautmann, C. H. & ORourke, T. D. (1985). Lateral force displacement response of buried pipe. J. Geotech. Engng ASCE 111, No. 9, 10681084. Trautmann, C. H., Kulhawy, F. H. & ORourke, T. D. (1985). Sand density measurements for laboratory studies. Geotech. Test. J. ASTM 6, No. 4, 159165. Trautmann, C. H., ORourke, T. D., Grigoriu, M. D. & Khater, M. M. (1986). Systems model for water supply following earthquakes. In Lifeline seismic risk analysis: case studies (ed. R. Eguchi), pp. 3050. Reston, VA: ASCE.

3418. Washington, DC: Minerals Management Service, US Department of Interior. Mitchell, J. K. & Soga, K. (2005). Fundamentals of soil behavior, 3rd edn. Hoboken, NJ: John Wiley & Sons. National Academy of Engineering and National Research Council (NAE/NRC) (2009). The New Orleans Hurricane Protection System: Assessing pre-Katrina vulnerability and improving mitigation and preparedness. Washington, DC: National Academies Press. National Research Council (NRC) (2002). Review procedures for water resources project planning. Washington, DC: National Academies Press. National Research Council (NRC) (2009). Final Report from the NRC Committee on the review of the Louisiana Coastal Protection and Restoration (LACPR) Program. Washington, DC: National Academies Press.. Nodine, M. C., Cheon, J. Y., Wright, S. G. & Gilbert, R. B. (2007). Mudslides during Hurricane Ivan and an assessment for the potential for future mudslides in the Gulf of Mexico. MMS Project No. 552, Phase II Project Report prepared for Minerals Management Service by University of Texas at Austin and Offshore Technology Research Center. Olson, N. A. (2009). Soil performance for large scale soil-pipeline tests. PhD thesis, Cornell University, Ithaca, NY. ORourke, M. J. & So, P. (2000). Seismic fragility curves for ongrade steel tanks. Earthquake Spectra 16, No. 4, 801815. ORourke, T. D. (1998). An overview of geotechnical and lifeline earthquake engineering, Geotechnical Special Publication No. 75, Vol. 2, 13921426. Reston, VA: ASCE. ORourke, T. D. (2007). Critical infrastructure, interdependencies, and resilience. The Bridge, Spring, 2229. ORourke, T. D. & Bonneau, A. (2007). Lifeline performance under extreme loading during earthquakes. In Earthquake geotechnical engineering (ed. K. D. Pitilakis), pp. 407432. Dordrecht: Springer. ORourke, T. D. & Pease, J. W. (1992). Large ground deformations and their effects on lifelines facilities: 1989 Loma Prieta earthquake, NCEER-92-0002, Vol. 2, pp. 5.15.85. Buffalo, NY: MCEER. ORourke, T. D. & Pease, J. W. (1997). Mapping liqueable layer thickness for seismic hazard assessment. J. Geotech. Engng ASCE 123, No. 1, 4656. ORourke, T. D. & Toprak, S. (1997). GIS assessment of water supply damage from the Northridge earthquake. In Spatial analysis in soil dynamics and earthquake engineering (ed. D. Frost), pp. 117131. Reston, VA: ASCE. ORourke, T. D., Beaujon, P. A. & Scawthorn, C. R. (1992). Large ground deformations and their effects on lifeline facilities: 1906 San Francisco earthquake, NCEER-92-0002, Vol. 2, pp. 1.1 1.130. Buffalo, NY: MCEER. ORourke, T. D., Stewart, H. E. & Jeon, S.-S. (2001). Geotechnical aspects of lifeline engineering. Proc. Instn Civ. Engrs Geotech. Engng 149, No. 1, 1326. ORourke, T. D., Wang, Y. & Shi, P. (2004). Advances in lifeline earthquake engineering. Proc 13th World Conf. on Earthquake Engng, Vancouver, paper no. 5003 (CD-ROM). ORourke, T. D., Bonneau, A., Pease, J., Shi, P. & Wang, Y. (2006). Liquefaction ground failures in San Francisco. Earthquake Spectra 22, No. 52, 60916112. ORourke, T. D., Jezerski, J. M., Olson, N. A., Bonneau, A. L., Palmer, M. C., Stewart, H. E., ORourke, M. J. & Abdoun, T. (2008). Geotechnics of pipeline system response to earthquakes. Proc. Geotechnical Earthquake Engineering and Soil Dynamics IV, Sacramento, CA, 138. Palmer, M. C., ORourke, T. D., Stewart, H. E., ORourke, M. J. & Symans, M. (2006). Large displacement soilstructure interaction test facility for lifelines. Proc. 8th US Nat. Conf. on Earthquake Engineering, San Francisco, CA, paper no. 1257 (CD-ROM). Pham, H. Q., Fredlund, D. G. & Barbour, S. L. (2005). A study of hysteresis models for soil water characteristic curves. Can. Geotech. J. 42, No. 6, 15481568. Rauber, R. M., Walsh, J. E. & Charlevoix, D. J. (2005). Severe and hazardous weather, 2nd edn. Dubuque, IW: Kendall/Hunt Publishing Co. Reuss, M. (2004). Designing the bayous: The control of water in

GEOHAZARDS AND LARGE, GEOGRAPHICALLY DISTRIBUTED SYSTEMS

Turban, E. (1995). Decision support and expert systems: Management support systems, 4th edn. Upper Saddle River, NJ: Prentice Hall. Turner, J. E. (2004). Lateral forcedisplacement behavior of pipes in partially saturated sand. MS thesis, Cornell University, Ithaca, NY. Ubilla, J., Abdoun, T., Sasanakul, I., Sharp, M., Steedman, S., Vanadit-Ellis, W. & Zimmie, T. (2008). New Orleans levee system performance during Hurricane Katrina: London Avenue and Orleans Canal south. J. Geotech. Geoenviron. Engng ASCE 134, No. 5, 668680. US Army Corps of Engineers (USACE) (2009). Louisiana coastal protection and restoration: Final Technical Report. US Army Corps of Engineers, New Orleans District (http://lacpr.usace.army.mil/default.aspx?pLACPR_Final_Technical_Report). US Census Bureau (2009). US Census Bureau news release, 1 July (http://www.census.gov/Press-Release/www/releases/archives/ population/013960.html). US Energy Information Administration (2005). Hurricane Katrina impacts on the US oil and natural gas markets: Special report, 31 August (http://tonto.eia.doe.gov/oog/special/eia1_katrina_ 083105.html/) US Geological Survey (USGS) (2008). The ShakeOut scenario, USGS Open File Report 2008-1150, CGS Preliminary Report 25. US Geological Survey, California Geological Survey (http:// pubs.usgs.gov/of/2008/1150/). Wang, Y. & ORourke, T. D. (2008). Seismic performance evaluation of water supply systems, MCEER-09-0008. Buffalo, NY: MCEER, University of New York at Buffalo. Whitman, R. V., Anagnos, T., Kircher, C. A., Lagorio, H. J., Lawson, R. S. & Schneider, P. (1997). Development of a national earthquake loss estimation methodology. Earthquake Spectra 13, No. 4, 643661. Wisch, D. J. & Ward, E. G. (2007). The legacy of the terrible triplets and standards of the future. Proc. Offshore Technol. Conf., Houston, TX, paper OTC 18979. Wooten, R. L., Gilbert, R. B., Marcuson, W. F., Harder, L. F. & Nicolson, P. G. (2009). Reconnaissance of the New Orleans hurricane and storm damage reduction system after Hurricane Gustav, Report No. GEER-015. Berkeley, CA: Geoengineering Extreme Events Reconnaissance Association. Yimsiri, S., Soga, K., Yoshizaki, K., Dasari, G. & ORourke, T. D. (2004). Lateral and upward soilpipeline interactions in sand for deep embedment conditions. J. Geotech. Geoenviron. Engng ASCE 130, No. 8, 830842. Youd, T. L. & Hoose, S. N. (1978). Historic ground failures in northern California triggered by earthquakes, USGS Professional Paper 993. Washington, DC: US Government Printing Ofce.

543

VOTE OF THANKS C. J. F. P. JONES, Emeritus Professor of Geotechnical Engineering, Newcastle University

Tom ORourke has been a regular visitor to the United Kingdom for some 30 years. During his early visits his interests were associated primarily with the effects of tunnelling on structures, and the effects of mining subsidence on buried services such as gas pipelines. The detrimental effects of mining subsidence on buried services, as well as on strategic structures such as bridges, is well understood in the coalelds of Britain, and Professor ORourkes studies were clearly aligned with those interested in these subjects, particularly in the North of England. Mining subsidence effects on buried services are an example of system components operating under extreme conditions of soilstructure interaction. In essence, mining subsidence has been described as being similar to a slow earthquake. Professor ORourke was always a popular lecturer during his frequent visits, with a natural ability to capture the audiences attention, as he has done this evening. He notably explained to engineering students that geotechnical engineering can be described by the three Ds: dirty, difcult and dangerous. Tom has now moved on from a study of individual system components, extending his experience and knowledge to the management of large geographical distribution systems that are affected by a range of potential geohazards such as earthquakes, oods and hurricanes. This work could be described by the three Is: intelligent; integrated and inuential. Intelligence comes from a fundamental understanding of complex soilstructure interaction, which offers improvements in our ability to model soilstructure behaviour. Integration results from the use of advanced geographical information systems and remote sensing, and permits this understanding to be applied to large distributed lifeline systems. The work is inuential in that it widens the role that geotechnical engineers have in society, whereby they now play a critical role in managing the performance of large infrastructure networks that are susceptible to natural hazards. In his lecture Professor ORourke has shared his unrivalled experience of the management systems resulting from his studies, illustrating their effectiveness in the management of the Los Angeles and San Francisco water distribution networks, and as a tool in understanding the effects of hurricanes on New Orleans and the Gulf of Mexico oil and gas networks. Tom ORourkes lecture looks to the future of geotechnical engineering, and points the way for engineers to take a more proactive and inuential role in the management of the infrastructure systems upon which society depends. It is with great pleasure that I invite the audience to thank Professor ORourke for an elegant and inuential Rankine lecture.