10
Recommendations

Major findings and conclusions of the committee that support the recommendations in this chapter are summarized in the “Key Findings and Conclusions” boxes at the beginning of each chapter. This chapter offers recommendations for improving engineering practice and scientific understanding in regard to the triggering and consequences of liquefaction based on an assessment of past and current research presented in the earlier chapters of this report. As such, recommendations in this report reinforce old ideas, expand or move beyond current practice, or offer new directions for research. Some of these recommendations can be implemented with existing tools and are intended to help geotechnical engineers make better decisions in daily practice. Others are directed at researchers and are intended to advance understanding of liquefaction-related phenomena so that current approaches to assessing liquefaction triggering and its consequences may be improved. Still other recommendations look beyond current approaches toward new assessment methods. Developing and implementing these methods will require collaboration among engineers, geologists, and others, and many would benefit from community-based efforts. These will certainly require some level of funding. The organization and funding of such efforts is beyond the scope of this report.

The recommendations are grouped by the main topics in the committee’s Statement of Task (see Box 1.4):

1. collecting, reporting, and assessing the sufficiency and quality of data;
2. addressing the spatial variability and uncertainty of these data; and
3. developing improved tools for assessing liquefaction triggering and its consequences.

Each boldfaced recommendation is followed by summaries of supporting evidence and, where appropriate, examples of the improvements or information needed.

COLLECTING, REPORTING, AND ASSESSING DATA SUFFICIENCY AND QUALITY

Recommendation 1. Establish curated, publicly accessible databases of relevant liquefaction triggering and consequence case history data. Include case histories in which soils interact with built structures. Document the case histories with relevant field, laboratory, and physical model data. Develop the databases with strict protocols and include indicators of data quality.

Field data from liquefaction case histories underpin the development, calibration, and validation of predictive models for liquefaction triggering and its consequences. For maximum benefit to the earthquake engineering community, case history databases need: (1) cases with parameter values beyond the ranges found in current databases; (2) updating to include the wealth of data available from recent earthquakes; (3) greater consistency with respect to information included in the databases and greater transparency about differences in quality, levels of detail, degree of vetting, and documentation; and (4) open access with capabilities for searching.

Large amounts of liquefaction case history data collected since 1995 are not currently included in the most widely used predictive models. Incorporating this information could enhance the completeness of the available field data with regard to such parameters as tectonic setting, earthquake magnitude, and soil type.

Current databases generally lack field case histories that represent behavior of liquefiable soils

• at depths greater than about 15 meters beneath the ground surface and initial vertical effective stresses larger than approximately 100 kPa (1 atm);
• subjected to relatively small and very large magnitude events (i.e., less than about M 5.9 and greater than approximately M 7.8);
• containing greater than 35% fines content;
• containing greater than 50% fines content and of low plasticity; and
• that are in sloping ground or are adjacent to free faces (i.e., may be subject to lateral spreading) and have normalized standard penetration test (SPT) values (N₁)₆₀ (or (N₁)_60-cs if adjusted clean-sand values are used) of greater than 15 blows per 30 cm or normalized cone penetration test (CPT) resistance values q_c1N (or q_c1Ncs if adjusted clean-sand values are used) greater than 85 atmospheres (8.6 MPa).

The databases are also unbalanced overall, being more heavily populated with cases where liquefaction has been observed than where it has not occurred. Case histories where potentially liquefiable soils have not liquefied can be as important as case histories where liquefaction has occurred, and they are particularly important in establishing limits for liquefaction triggering and its consequences with regard to magnitude, depth, initial static shear stress, grain size, and plasticity. Future field studies should take particular notice of these data gaps and strive to fill them with case histories that would influence the boundary between liquefaction and non-liquefaction.

Current databases of consequences of liquefaction—especially of the residual shear strength of liquefied soil, of lateral spreading displacements, and of the interaction between liquefied soils and man-made structures—also need to be strengthened. This is particularly true for intermediate density and dense soils where liquefaction effects may be subtle or possibly inconsequential: namely, soils with a normalized CPT tip resistance, q_c1-cs, greater than about 85 atm (8.6 MPa); standardized and normalized SPT resistance, N_1,60-cs, greater than approximately 15 blows per 30 cm; or a normalized shear wave velocity, V_S1, greater than about 225 m/s. Collecting data from sites that meet these criteria and that have been subjected to strong ground motions, even if there are no observed effects of liquefaction, is a means to fill this data gap.

Few compilations of case histories document soil-structure interaction effects at sites that have liquefied. Protocols should be developed for collecting data on soil-structure interaction effects (e.g., the impact of lateral spreading on deep foundations and buried pipelines, bearing failure of shallow foundations, the impact of preexisting structures on the amount of lateral spreading), and more attention should be paid to collecting this type of data.

Relevant laboratory and physical model data can both augment field data in publicly accessible, high-quality databases and help improve prediction models for conditions not adequately constrained by case history data. Strict protocols for data included in the databases, such as those being developed for the Next-Generation Liquefaction project (see Chapter 3, Box 3.3), need to be established and followed to ensure consistency of quality. Implementation of Recommendation 1 cannot be fully realized without an adequately funded and coordinated effort.

Recommendation 2. Characterize locations with high probability of liquefaction and establish them as field observatories for liquefaction triggering and its consequences.

Important gaps in case history databases could be filled by establishing liquefaction observatories at sites that are well characterized, well instrumented, and strategically located in areas where there is a high probability of earthquake-induced soil liquefaction in coming decades. Detailed high-quality characterization of the pre-earthquake surface and subsurface conditions at field observatories that then capture liquefaction events would strengthen the value of recorded data by removing some uncertainty related to pre-liquefaction soil properties and liquefaction-induced effects (e.g., changes in soil properties and ground displacements). Data from these observatories can be used to calibrate and validate procedures for assessment of liquefaction triggering and its consequences. Properly instrumented observatories may help fill important data gaps, including behaviors associated with liquefaction at depth, liquefaction of gravels and fine-grained soils, liquefaction on sloping ground and within and beneath embankments, and soil-structure interaction effects in liquefied soil. Information from such observatories could inform community-based collaborative efforts that will allow a more complete understanding of liquefaction triggering and consequences.

Characterization data to be collected at observatory sites, where feasible, should include the sedimentary and seismological history of the site, modifications made by humans, results of both in situ and laboratory testing, lateral and vertical variability in liquefiable soils, and surface and subsurface evidence of previous liquefaction. Site characterization also needs to include the collection of high-quality geographically referenced and orthorectified aerial and satellite images as well as high-precision topographic data (e.g., from Light Detection and Ranging surveys). These data need to be updated periodically to account for geomorphic or anthropogenic changes that may occur prior to recorded earthquake activity. Instrumentation should include strong motion instruments at the surface and at depth (ideally above and below liquefiable layers) and pore-pressure transducers within liquefiable soil strata and at other strategic locations in the soil profile; inclinometers in areas subject to potential lateral spreading; settlement probes; and appropriate instrumentation (e.g., strong motion recorders, strain gages, inclinometers) of structural elements of on-site facilities (e.g., piles, foundation slabs, building superstructures, and bridge decks).

Several free-field borehole arrays in the United States and elsewhere (see Box 3.3) meet the minimum requirements for observatories for evaluation of both empirical and numerical liquefaction triggering models. More sites with this type of instrumentation are needed to account for the diverse geologic, seismologic, and geotechnical conditions where liquefaction is possible. Given the low probability of a triggering event at any specific site, having multiple instrumented sites will increase the likelihood of capturing a triggering event.

Few case histories that include detailed measurements of soil-foundation-structure interaction effects on liquefiable ground are available. More sites at which instrumentation is installed to capture such soil-structure interaction effects in liquefied soils are needed. A cost-effective

technique to obtain useful data may be to place new liquefaction field observatories at already well-characterized sites (e.g., sites instrumented as part of a strong motion instrumentation program). There are challenges associated with implementing and funding such a program, but the California Strong Motion Instrumentation Program,¹ established primarily to provide data on the seismic response of structures, is one model of how such observatories can be funded and maintained.

Recommendation 3. Use data from the cone penetration test (CPT) for field-based estimates of liquefaction resistance where feasible. If the standard penetration test (SPT) is used for this purpose, make hammer energy measurements. Supplement field-based estimates with other methods as appropriate to characterize the site.

CPT soundings offer advantages over other methods of estimating liquefaction resistance in detecting thin layers that may affect liquefaction triggering and subsequent pore-pressure redistribution. CPT results are less dependent on the equipment operator or setup than most other in situ test methods, and the CPT can be performed with relative speed and economy. The CPT can also provide a cost-effective means to measure shear wave velocity (VS) that can enhance site characterization and liquefaction assessment. CPT equipment is available that can collect samples, but this method is not generally available, and the technology at the time of this writing is not mature. Using seismic piezocone penetration testing (CPTu) should be considered because the additional information from the CPTu can provide insight to soil layering, soil dilatancy, and groundwater conditions, and the additional information from the seismic cone can be used for other purposes (e.g., soil layering, site response analyses, and as an alternative method to evaluate liquefaction triggering). There are sites, however, where CPT soundings are not feasible, such as those with gravelly or very dense soils. In such cases, other techniques for assessing the liquefaction resistance of the soil, including the SPT, alternative means to collect V_s measurements, the Becker Penetration Test (BPT), and large diameter dynamic cone penetration testing, have proven useful. If the BPT is used, an instrumented BPT rig in which energy delivered to the sampler is measured (i.e., the BPTi), should be used.

The SPT is still widely used in practice and has a well-defined role in cases where direct measurement of fines content is needed or is desirable to reduce uncertainties (e.g., for interpretation of CPT and V_s data). SPT blow counts may also be useful as a supplemental technique for assessing soil resistance in dense soils and at depths where the CPT cannot penetrate. Nonetheless, there are several operational factors for which corrections are difficult to make. The most variable of these, and that which often requires the largest correction factor, is hammer energy. If the SPT blow count is to be used to assess resistance of a soil to liquefaction, SPT hammer energy measurements are needed to reduce the uncertainty associated with SPT blow count values. The SPT setup should conform to conditions for which no correction factor is needed (e.g., standard borehole diameter, sampler configuration). Rotary wash borings for SPT testing are more reliable than are hollow stem auger borings, as SPT blow counts recorded in hollow stem auger borings below the water table are particularly susceptible to error and should be carefully evaluated for indications of borehole disturbance (i.e., abnormally low blow count values).

Use of multiple methods for any assessment may help to constrain uncertainties. The source of any discrepancies in assessments should be identified to inform professional judgement of the

___________________

¹ See www.consrv.ca.gov/cgs/smip.

design engineer and to allow the engineer to decide which methods are most reliable in the given circumstances.

Recommendation 4. When refining or developing new empirical relationships for use in liquefaction analyses, incorporate unbiased estimates for input parameters; identify and quantify when possible the uncertainty associated with those estimates; and use soil mechanics principles, seismologic principles, and experimental data to extrapolate beyond ranges in which field data constrain the empirical relationships.

Relationships for analysis of liquefaction triggering and consequences based on case history data include cyclic resistance curves developed for simplified stress-based analysis of triggering, relationships between post-liquefaction shear strength and penetration resistance, and equations for evaluating the levels of liquefaction-related lateral spreading. Empirical procedures generally yield consistent estimates of liquefaction hazards for conditions within the ranges that are constrained by the field data. Databases of field case histories, however, do not include data for the full range of conditions over which a liquefaction assessment may be required (e.g., depths of about 15 m or greater, gravelly or silty soils, earthquakes larger than magnitude 8, presence of static shear stress from a topographic slope). Empirical procedures that employ a polynomial functional form based solely on the best statistical fit to available data may yield unrealistic results when used to make predictions beyond the limits of the data. In contrast (as discussed in Chapter 4), extrapolation using functional forms based on experimental data trends (e.g., from simple shear and triaxial tests and from centrifuge testing) and on principles of soil mechanics, wave propagation theory, or seismology, as appropriate, provide a more appropriate means of extrapolating beyond the limits of the field data. Developers of empirical procedures need to state clearly the range over which their procedures are constrained by the field data and the method used to extrapolate beyond that range.

Those developing or refining empirical relationships should attempt to use approaches that minimize bias, rather than simplified equations with a built-in bias, when quantifying the adjustment factors, correlations, and parameter relationships used in the relationship. Developers also need to identify and, if possible, to quantify the associated uncertainty with each adjustment factor, empirical correlation, and parameter relationship. This will avoid the compounding of uncertainties and will facilitate assessment of the overall uncertainty associated with the method. Employing adjustment factors, empirical correlations, and parameter relationships with built-in bias when developing a liquefaction triggering or consequence assessment method compounds uncertainties and makes it difficult, if not impossible, to assess the overall uncertainty associated with the method—an important step in liquefaction assessments. Assessing the overall uncertainty is an important consideration when recommending a factor of safety (FS) to use in a liquefaction triggering or consequence assessment.

ADDRESSING THE SPATIAL VARIABILITY AND UNCERTAINTY OF DATA

Recommendation 5. Use geology to improve the geotechnical understanding of case histories and project sites, particularly where potentially liquefiable soils vary in thickness, continuity, and engineering properties.

Current site assessment practice tends to focus on the engineering characteristics of subsurface materials. Their geologic context, however, is basic to assessing liquefaction hazards both in case histories and in project work. Opportunities to use geology more effectively include incorporating descriptions of sedimentary geology in protocols for site assessment of case histories and in site characterization for project work. In case histories, inferences about which deposits liquefied, and which did not, could include characterization of the vertical and lateral variability of postulated critical layers (deposits inferred to have liquefied) and of subsurface paths of fluid escape that dikes and sills can mark. Such geologic investigations and characterizations could aid interpretation of case histories where the sedimentary geometry is complex, as is commonly the case on alluvial fans and in river meander belts, and they could be extended to project work where detailed understanding of liquefaction potential is warranted by threats to life and property. Understanding the lateral continuity and variability of liquefiable layers can be particularly important in assessing the potential for lateral spreading.

Geologic understanding of surficial deposits also underpins regional maps of liquefaction hazard that can serve as screening tools in project work. These regional maps—sometimes prepared by state agencies—are becoming increasingly quantitative through studies that relate surficial geologic units to probabilistic estimates of liquefaction potential. Further improvements to the maps could include representing subsurface stratigraphy where liquefaction may occur in stratigraphic units that widely underlie those mapped at the surface.

A database of geologic evidence for liquefaction in the central and eastern United States could be expanded worldwide, and it could provide information on the sedimentary environment, depositional age, and tectonic setting to guide evaluation of the influence of these factors on liquefaction potential. This effort, in turn, might enable case history databases to be made more complete by including data on geologic evidence for liquefaction in earthquakes in the 19th century and before. Extending temporal coverage of liquefaction case histories in this manner could be particularly helpful in areas where major earthquakes recur at intervals of hundreds or thousands of years and where there are few case histories of any age.

These suggested efforts are meant to strengthen use of geology in assessments of liquefaction triggering and consequences. They would supplement current best practices that include accurate description, sound interpretation, and qualified professional judgment regarding tectonics, landforms, regional geology, sedimentology, groundwater hydrology, and past performance of analogous sites.

Recommendation 6. Implement simplified stress-based methods for liquefaction triggering in a manner consistent with how they were developed. Avoid using techniques and adjustment factors from one variant of a method with other variants. Consider using more than one simplified method when making a liquefaction triggering assessment.

The simplified stress-based approach to liquefaction triggering analysis has served the profession for more than 40 years and, despite shortcomings, will continue to be a mainstay of earthquake engineering practice for the near future. Several variants of the simplified stress-based method have been developed. Each variant has associated adjustment factors, empirical correlations, and parameter relationships that are used during its implementation. If the adjustment factors, empirical correlations, and parameter relationships from one variant of the simplified stress-based procedure are applied to another variant (e.g., the mixing and matching of factors,

correlations, and relationships) errors are likely to be introduced, and both bias and uncertainty will be added to the liquefaction triggering assessment.

Each of the variants of the simplified stress-based method, for example, includes a means of assessing the earthquake-induced shear stress with depth as the stress reduction coefficient, rd. The location of the triggering curve in a particular variant of the simplified method depends upon the method used to assess rd. Therefore, using rd from one variant of the simplified method (e.g., the Idriss and Boulanger, 2008, variant) in conjunction with the triggering curve from another variant (e.g., the Cetin et al., 2004, variant) will introduce unquantified (and unquantifiable) errors into the triggering assessment.

Similarly, it is tempting to assume that, rather than using a simplified equation for rd to assess the earthquake-induced shear stress, using a more precise analytical method for assessing earthquake-induced shear stress (e.g., a site response analysis) will increase the accuracy of the analysis. In instances where the simplified equation for rd incorporates biased estimates of earthquake-induced shear stress (see, e.g., Idriss and Boulanger, 2008), however, using a different method to compute the shear stresses will introduce additional and unquantifiable uncertainty into the liquefaction assessment.

There are significant epistemic (modeling) uncertainties associated with all variants of the simplified stress-based approach to liquefaction triggering assessment. Situations may warrant the use of more than one variant, such as where analysis results indicate a marginal FS against triggering, and where the consequences of triggering are considered to be unacceptable (e.g., require costly remediation or threaten life safety). A more sophisticated liquefaction assessment may be warranted if the results of the triggering assessment in these situations are contradictory or marginal. Multiple variants of the simplified method are also warranted when the method is applied beyond the bounds where method relationships are constrained by field data (see Recommendation 1 for these limits).

As stated previously, using multiple methods may help constrain uncertainties among test methods. The sources of discrepancies among test method results should be identified. It is then up to the professional judgment of the investigator to determine how results should be weighed, or whether the use of more advanced methods may be warranted.

Recommendation 7. In developing methods to evaluate liquefaction triggering and its consequences, explicitly incorporate uncertainties from field investigations, laboratory testing, numerical modeling, and the impact of the local site conditions on the earthquake ground motions.

It is especially important for both the developer of methods and the practicing engineer to consider uncertainty in procedures used for assessment of liquefaction triggering and its consequences. All aspects of liquefaction assessment—from characterizing the site and site-specific ground motions to assessing the severity of liquefaction consequences—are fraught with uncertainty. The uncertainties in field data and analytical and experimental results need to be stated in the form of error bounds, standard deviations, bias, or other statistically appropriate measures.

Most field investigations, laboratory test programs, and analysis procedures do not explicitly address uncertainty. Descriptions of procedures to evaluate liquefaction do not always state clearly where and what the uncertainties are and typically do not address how to incorporate them into the liquefaction assessment. Furthermore, while many factors used for liquefaction assessment require correction, the correction factors (e.g., corrections applied to SPT blow counts, magnitude scaling

factor) are themselves uncertain. The uncertainty in the correction factor is rarely accounted for, and in some cases, it may introduce even more uncertainty in the results of the analysis than that associated with the original measurement (see Box 4.3).

Uncertainty in a liquefaction consequence or triggering assessment increases rapidly as the assessment method moves beyond the range within which it is constrained by field data. Because the data ranges over which many assessment procedures are applicable may not be adequately identified, users of the procedures may be tempted to extrapolate beyond the ranges for which the procedures are valid without considering the increased uncertainty. Good engineering practice dictates that any extrapolation beyond existing data be supported with lab testing or other sources of information. Liquefaction assessment procedures need to clearly identify the range over which they are constrained by both field and laboratory test data. Users need either to refrain from extrapolating beyond this range or to explicitly identify when they have done so and qualify the results accordingly.

Recommendation 8. Refine, develop, and implement performance-based approaches to evaluating liquefaction, including triggering, the geotechnical consequence of triggering, structural damage, and economic loss models to facilitate performance-based evaluation and design.

Earthquake engineering is, in general, moving toward performance-based design, and liquefaction issues will need to be addressed within that framework. Regional variations in seismicity, coupled with the significant uncertainties in earthquake loading and liquefaction resistance, warrant probabilistic characterization for the evaluation of both liquefaction triggering potential and the consequences of liquefaction. Quantifying the uncertainties associated with existing and new procedures is necessary to facilitate risk-based approaches to liquefaction triggering and consequence assessment. Geotechnical engineers need to acquire and document data to support quantification of uncertainties and to develop efficient computational tools that make risk-based calculations practical.

Probabilistic liquefaction triggering and consequence evaluation procedures can be convolved with the results of a probabilistic seismic hazard analysis (PSHA) to produce fully probabilistic liquefaction hazard analyses (PLHA). A PLHA considers all levels of shaking (rather than just that associated with a single return period), all contributing magnitudes (rather than just the mean or modal magnitude) at all return periods, and the uncertainty in ground motion at all levels to provide a complete and consistent indication of liquefaction hazards across different seismic environments. To accomplish this, computational tools need to be integrated with PSHA tools, making certain they allow weighted contributions from multiple predictive models to account for epistemic uncertainty.

The geotechnical community needs to develop PLHA to meet the growing demand from the greater earthquake engineering community for risk-based liquefaction assessment as part of the trend toward performance-based design. The results of a PLHA should be expressed in the form of a response curve that indicates how often different levels of response (e.g., liquefaction triggering, lateral spreading displacement, settlement) can be expected to occur. In a performance-based framework, the results of this type of PLHA can be convolved with damage and loss models to estimate the risk associated with liquefaction. Risk can be expressed as a risk curve that indicates how often different levels of loss can be expected to occur. Losses include loss of life, loss of functionality, and direct and indirect economic consequences.

Implementation of performance-based design often requires the use of advanced computational methods. Keeping in mind the need to use more than one strong ground motion time history, the amount of computation required for even a single time history analysis, and the need to conduct sensitivity analyses regarding modeling assumptions, a balance must be struck between the complexity of the numerical model, the detail to which the site is or can be characterized, and computational efficiency. The balance to be struck depends on the specifics of the problem: for instance, what is of primary concern. No general guidance can be provided on how to strike this balance; therefore, the site investigation and numerical analysis plans and analyses results should be peer reviewed by an engineer experienced in this type of analysis, as recommended by the ASCE (2010) for site-specific seismic hazard assessments.

IMPROVING TOOLS FOR ASSESSING LIQUEFACTION TRIGGERING AND ITS CONSEQUENCES

Recommendation 9. Use experimental data and fundamental principles of seismology, geology, geotechnical engineering, and engineering mechanics to develop new analytical techniques, screening tools, and models to assess liquefaction triggering and post-liquefaction consequences.

The simplified methods for liquefaction triggering and consequence assessment that are typically employed in practice today are limited in their predictive capabilities. This limitation also exists with the more sophisticated empirical and analytical methods used for liquefaction assessment. Furthermore, most available methods for assessment of liquefaction triggering and consequence assessment are not compatible with probabilistic characterization of the seismic loading that has become standard in engineering practice. New approaches to liquefaction assessment are needed that would go beyond improving empirical methods (Recommendation 4). Ideally, these new approaches will:

1. relate quantitatively liquefaction hazards to geologic units at appropriate depths;
2. increase predictive capabilities in both evaluation of triggering and consequences;
3. be compatible with probabilistic characterization of seismic loading;
4. be consistent with performance-based assessments of liquefaction consequences; and
5. be readily accessible to practicing engineers.

To relate liquefaction hazards quantitatively to geologic units at appropriate depths, geotechnical engineers will have to work with engineering geologists to characterize the depth-dependent distribution and properties of geologic units. To increase the accuracy of predictions, new approaches need to be based on sound science (e.g., geology, seismology, and physics) and fundamental principles of engineering mechanics and geotechnical engineering. The approaches need to be consistent with patterns of behavior identified through analysis of field case histories and laboratory and physical model test data. Strain-based and energy-based approaches are two avenues of development that merit consideration.

A next logical step in model development is to introduce approaches compatible with probabilistic seismic hazard characterization that can include or be integrated with liquefaction consequence assessment. Rigorous validation of these models using case history data and development of implementation approaches that make the models accessible to practitioners is

needed to facilitate their adoption in engineering practice. Development and adoption in practice of such integrated and validated models would be consistent with trends in other areas of modern earthquake engineering practice.

Recommendation 10. Develop and validate computational models for liquefaction analyses. Use laboratory and physical model tests at different spatial scales and case histories to provide insight into fundamental soil behavior and to validate the application of constitutive models to boundary-value problems.

While constitutive and computational models are intimately linked, computational models do not always accurately reproduce behavior across laboratory-test, physical-test, and field scales. Some limitations of computational modeling are related to the inability of the models to capture complex material behavior under cyclic loading, including the intimately coupled deformation and flow patterns before and after triggering. Such patterns may induce changes to solid-to-viscous-liquid behavior transformations, localized and large deformations, and porosity redistribution in the ground, all of which can only be predicted faithfully with improved computational models. Models based on discrete element mechanics offer a promising avenue for modeling mechanisms at the grain level, but they remain computationally burdensome. Further developments are needed to enhance the accuracy and computational efficiency of discrete methods and to couple the discrete methods with simulations across scales. Mesh-free methods, such as the material point method, offer great potential for modeling large strain flow behavior. Rigorous validation procedures that distinguish between accuracy of the computational method and the complexity of the problems for which the predictive model has been validated are required for all computational procedures, regardless of the computational approach and method of analysis they employ.

Recommendation 11. Conduct fundamental research on the stress, strain, and strength behaviors of soils prior to and after liquefaction triggering; devise new laboratory and physical model experimental techniques to aid development of constitutive models of those behaviors.

Fundamental research on liquefaction-related phenomena, both theoretical and experimental, is still required to advance the state of the art and practice. The stress-strain behavior of soils prior to and following liquefaction triggering is complex and still not fully understood. No available constitutive model captures all of the relevant soil behavior within a rigorous physics-based framework. In addition to changes in stiffness and strength of the soil caused by earthquake-induced excess porewater pressure, there are also changes in soil fabric and porosity. Needed research in these areas includes research on dilation that occurs in soils prior to and after liquefaction triggering, the solid–to-viscous-liquid behavior transformation that can accompany triggering, and the mechanisms that control the post-triggering fabric degradation that influences stress-strain behavior of soils. In particular, the mechanisms of post-triggering particle rearrangement and reconsolidation and the effects of relative density and consolidation stress on these mechanisms are not well understood and require further research. Understanding the mechanisms of post-liquefaction soil behavior is fundamental to understanding the dependency of post-liquefaction residual strength on consolidation stress; this includes the potential for a threshold value of soil resistance (e.g., normalized SPT blow count or CPT tip resistance) above

which residual shear strength increases so precipitously that flow sliding and lateral spreading are not of concern.

IMPROVING RESEARCH AND PRACTICE

Today’s empirical approaches to liquefaction assessment reflect a half century of development and refinement. Society has benefited from widespread acceptance and application of these approaches, but now those approaches to liquefaction assessment are changing in response to new data, new analytical methods, and recent developments in earthquake engineering. The preceding recommendations stress the importance of implementing existing methods in accordance with their stated procedures, using high-quality data that depict spatial variations at a site, and making that high-quality data available publicly after careful review and vetting. The recommendations also emphasize the development of fully PLHAs that can be used in a performance-based design framework. In the meantime, predictive models and tools consistent with fundamental principles of dynamics and soil mechanics need to be developed to meet the need of extrapolating beyond the ranges of existing data.

Advances such as these will require a concerted effort not only by researchers and practitioners but also by facility owners and stakeholders at large. Researchers need to continue to collect high-quality field case history data supplemented with laboratory and model test data as well as to develop and validate new, more accurate models to assess earthquake-induced soil liquefaction. Practitioners need to make greater efforts to understand the limitations of their choices of characterization and assessment approaches, and they need to better understand the specific geologic and tectonic controls on liquefaction hazards at their sites. Facility owners need to contribute to this effort by supporting the establishment of public case history databases and well-characterized and instrumented field observatories for study of liquefaction effects.

Perhaps most importantly, liquefaction assessment needs to expand beyond empirically based methods with new methods that are informed by more data, by quantification of uncertainties, and by fundamental scientific and engineering principles. To be adopted by the engineering community, any new method needs to be responsive to practitioner needs. In particular, the

1. fundamental basis and applicability of the model needs to be understandable;
2. outputs from the model need to be translatable to the problems of interest;
3. limitations of the model need to be clearly defined;
4. model needs to be readily implementable by practicing engineers;
5. methods must be economical (i.e., not require excessive effort); and
6. methods must keep pace with the evolution of earthquake engineering practice toward performance-based design.

Moving along these recommended paths forward will fundamentally improve prediction of earthquake-induced soil liquefaction and its consequences. Reliable predictions will provide greater protection of life safety, of built infrastructure, and of the economy by reducing adverse economic, environmental, and social impacts of liquefaction.

__________________