Regional efficacy of a global geospatial liquefaction model

doi:10.1016/j.enggeo.2020.105644

Engineering Geology

Volume 272, July 2020, 105644

https://doi.org/10.1016/j.enggeo.2020.105644 Get rights and content

Highlights

•
The GGLM developed by Zhu et al. (2017) was modified by adding 0.1 g PGA and 1700 mm precipitation thresholds.
•
Adding these two thresholds improved the regional efficacy of the Global Geospatial Liquefaction Model (GGLM).
•
The liquefaction intensity index (LII) is developed to provide an event-level intensity measure for liquefaction.
•
The LII is based on observed severity of liquefaction in the field and the extent of liquefaction predicted by the GGLM.

Abstract

Liquefaction hazard maps are important for both pre- and post-event planning and mitigation. The global geospatial liquefaction model (GGLM) proposed by Zhu et al. (2017) and recommended for global application results in a liquefaction probability that can be interpreted as liquefaction spatial extent (LSE). The GGLM uses ShakeMap's PGV, topography-based Vs₃₀, distance to water body, water table depth, and annual precipitation as explanatory variables. The GGLM was originally developed and validated across 23 global earthquakes with most of the earthquakes in coastal settings. In this paper, LSE maps have been generated for 29 earthquakes around the world in a wide range of settings in addition to 23 of the original events to evaluate the generality and regional efficacy of the model. The GGLM was found to overpredict liquefaction spatial extent for earthquakes with large areas experiencing low PGA (below 0.1 g) and as a result, the GGLM has been modified to decrease over-prediction with the addition of a PGA threshold (no liquefaction when PGA < 0.1 g). The liquefaction intensity index (LII) is generated for each earthquake through the summation of LSE values across the event and compared with the liquefaction intensity inferred from the reconnaissance report. Using the LII as an indication of goodness of fit, the GGLM performs well across all regions with 5 earthquakes showing underprediction of LII and 4 earthquakes showing overprediction. When annual precipitation of a region exceeds 1700 mm, which is the upper quartile of annual precipitation in the development database, overprediction of liquefaction spatial extent is likely, and the use of threshold is recommended.

Introduction

Loosely deposited, cohesionless, and saturated soils may liquefy during cyclic loading from a major earthquake. Liquefaction may in turn induce ground failures of varying severity and result in damage to the built environment (Papathanassiou et al., 2015; Ocakoğlu and Akkiraz, 2019; Pokhrel et al., 2013; Jha and Suzuki, 2009). Soil vulnerability to liquefaction depends on soil density, water table depth (soil saturation) and earthquake demand (dynamic load). These factors are known to be dependent on geological age, depositional environment, topography, distance to water body and regional seismicity. Liquefaction assessment is an important component of an earthquake risk and loss evaluation for use in pre- and post-event planning and mitigation (Rahman et al., 2015; Chen et al., 2019; Rashidian and Gillins, 2018). Zhu et al., 2015, Zhu et al., 2017 developed a globally-applicable geospatial liquefaction model that predicts probability and spatial extent of liquefaction for use in developing liquefaction hazard maps for use in rapid response and loss estimation when local geotechnical data and surficial geologic maps are not available.

The Zhu et al., 2015, Zhu et al., 2017 geospatial liquefaction models (GLM) use globally available geospatial explanatory variables that are proxies for soil density, soil saturation and dynamic loading. The models were developed using logistic regression. The first Zhu et al. (2015) model used four earthquakes from New Zealand and Japan, in which liquefaction observations were spatially complete across each region (both liquefaction and non-liquefaction were mapped as continuous polygons). In the second Zhu et al. (2017) model, the GLM was updated by adding 22 earthquakes from China, Taiwan, Japan and the United States; however, the additional earthquakes did not have spatially complete maps of observed liquefaction; instead, liquefaction and non-liquefaction occurrences were primarily available as point data. The Zhu et al. (2017) publication proposed two alternate GLMs: one for coastal earthquakes (according to Zhu et al. (2017): “events where the liquefaction occurrences are, on average, within 20 km of the coast; or, for earthquakes with insignificant or no liquefaction, epicentral distances is less than 50 km to the coast”) and one for non-coastal earthquakes. Eighteen geospatial features were compared as proxies for soil density, soil saturation and dynamic loading of earthquake. The coastal GLM used PGV, Vs₃₀, distance to coast, distance to river and annual precipitation as the explanatory variables. The global (non-coastal) GLM used PGV, Vs₃₀, distance to water body (defined as minimum of distance to coast and distance to river), ground water table depth and annual precipitation as the explanatory variables. Zhu et al. (2017) also proposed a logistic function to translate the GLM liquefaction probability to an estimate of the liquefaction spatial extent (LSE). This allows the resulting map to provide an estimate of the fractional area of liquefaction within a pixel/polygon, which can be more directly compared with ground observations of liquefaction.

This study has two primary goals: 1) to test the global geospatial liquefaction model (GGLM) proposed by Zhu et.al (2017) on 52 earthquakes (including 29 new events not included in Zhu et al. (2017) and chosen to provide validation across the globe including regions such as Europe, Central America, and South America, which were not well represented in the model development) across the world to evaluate the model performance and regional efficacy; and 2) to enhance the GGLM's integration into an earthquake loss estimation system such as U.S. Geological Survey's Prompt Assessment of Global Earthquakes for Response (PAGER) system (Wald et al., 2010) for evaluating the relative intensity of liquefaction across earthquakes through the development of a liquefaction intensity index (LII). The model performance has been validated using observed liquefied areas reported by reconnaissance reports. Liquefaction observation after each of the 52 earthquakes used herein have been classified into 4 intensity categories; then, the GGLM-based LSE values are summed across the event and compared with the liquefaction intensity classes derived from these field observation reports to create an appropriate GGLM-based LII framework. Maps of LSE after an earthquake provide a visual of spatial extent and measure liquefaction intensity across the affected area; however, LSE does not provide an event-level intensity. As compared to ground shaking intensity, LSE is parallel to modified Mercalli intensity or PGA; whereas, the LII developed in this paper can be easily reported as a single event-level intensity measure.

Section snippets

Database and methodology

Zhu et al. (2017) presented coastal and global geospatial models for predicting liquefaction probability. While the coastal model relies on the distance to coast as a proxy for saturation and soil density, the global model uses distance to the nearest water and water table depth as the saturation proxies. As a result, the global model is recommended for global implementation and is preferred by the USGS (Thompson, personal communication, 2017) whereas the coastal model might be recommended for

Development of a GGLM-based liquefaction intensity index

In order to develop a GGLM-based liquefaction intensity index (LII_GGLM) to facilitate earthquake rapid response and loss estimation, each event in this study was evaluated in terms of liquefaction severity according to the index scale in Table 3. The field reconnaissance report for each event (referenced accordingly in Table 2) was carefully inspected to infer the severity of earthquake-induced liquefaction in terms of a qualitative rating (LII _{reconnaissance}) as presented in Table 3. The field

Conclusion and key remarks

The Global Geospatial Liquefaction Model (GGLM) proposed by Zhu et al. (2017) is evaluated for 52 coastal and non-coastal earthquake events in six different regions across the globe. GGLM was modified to include a PGA threshold of 0.1 g in addition to the original PGV threshold of 3 cm/s. The liquefaction spatial extent (LSE) derived from GGLM was summed (summation of liquefied area within individual pixels for the whole affected area) for each earthquake to calculate the total LSE (TLSE in km²

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgement

This work was supported under the U.S. Geologic Survey (USGS), Department of the Interior, under USGS Award Number G16AP00014; we gratefully acknowledge this support. The work has greatly benefited from discussions with Eric Thompson at USGS. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

References (65)

S. Bhattacharya et al.
Liquefaction of soil in the Tokyo Bay area from the 2011 Tohoku (Japan) earthquake
Soil Dyn. Earthq. Eng.
(2011)
J. Chen et al.
Toward a numerical-simulation-based liquefaction hazard assessment for urban regions using high-performance computing
Eng. Geol.
(2019)
M. Hamada et al.
Liquefaction-induced ground displacement and its related damage to lifeline facilities
Soils Found.
(1996)
K. Ishihara et al.
Case studies of liquefaction in the 1964 Niigata earthquake
Soils Found.
(1981)
S.K. Jha et al.
Liquefaction potential index considering parameter uncertainties
Eng. Geol.
(2009)
T. Mukunoki et al.
Reconnaissance report on geotechnical damage caused by an earthquake with JMA seismic intensity 7 twice in 28h, Kumamoto, Japan
Soils Found.
(2016)
F. Ocakoğlu et al.
A lower Pleistocene to Holocene terrestrial record from the Eskişehir Graben (Central Anatolia): Paleoclimatic and morphotectonic implications
Quat. Int.
(2019)
G. Papathanassiou et al.
Assessment of liquefaction potential for two liquefaction prone areas considering the May 20, 2012 Emilia (Italy) earthquake
Eng. Geol.
(2015)
G. Papathanassiou et al.
Recurrent liquefaction-induced failures triggered by 2014 Cephalonia, Greece earthquakes: spatial distribution and quantitative analysis of liquefaction potential
Eng. Geol.
(2016)
R.M. Pokhrel et al.
A kriging method of interpolation used to map liquefaction potential over alluvial ground
Eng. Geol.
(2013)

M.Z. Rahman et al.

Liquefaction hazard mapping by liquefaction potential index for Dhaka City, Bangladesh

Eng. Geol.

(2015)

V. Rashidian et al.

Modification of the liquefaction potential index to consider the topography in Christchurch, New Zealand

Eng. Geol.

(2018)

E.M. Rathje et al.

Coastal failures during the 1999 Kocaeli earthquake in Turkey

Soil Dyn. Earthq. Eng.

(2004)

I. Tohno et al.

Liquefaction of the ground during the 1978 Miyagiken-Oki earthquake

Soils Found.

(1981)

H. Yuan et al.

Liquefaction-induced ground failure: a study of the Chi-Chi earthquake cases

Eng. Geol.

(2004)

J.P. Bardet et al.

Initial Geotechnical Observations of the Bhuj, India, Earthquake of January 26, 2001

(2001)

J.D. Bray et al.

Some Observations of the Geotechnical Aspects of the February 28, 2001, Nisqually Earthquake in Olympia, South Seattle, and Tacoma, Washington

(2001)

J. Bray et al.

Geotechnical Engineering Reconnaissance of the August 24, 2014 M6 South Napa Earthquake

(2014)

G. Candia et al.

Geotechnical Reconnaissance of the 2015 Illapel Earthquake

(2015)

M. Carter et al.

Geotechnical Quick Report on the Affected Region of the 23 August 2011 M5. 8 Central Virginia Earthquake near Mineral, Virginia

(2011)

A.F. Chleborad et al.

Ground Failure Associated with the Puget Sound Region Earthquakes of April 13, 1949, and April 29, 1965 (No. 90–687).

(1990)

P. Clayton et al.

The Geotechnical Aspects of the September 3, 2016 M 5.8 Pawnee, Oklahoma Earthquake

(2016)

M. Cubrinovski et al.

Geotechnical Reconnaissance of the 2010 Darfield (New Zealand) Earthquake

(2010)

M. Cubrinovski et al.

Geotechnical Reconnaissance of the 2011 Christchurch, New Zealand Earthquake

(2011)

M. Erdik et al.

23 October 2011 Van (Turkey) earthquake

Nat. Hazards

(2012)

Y. Fan et al.

Global patterns of groundwater table depth

Science

(2013)

J.C. Gómez et al.

Soil liquefaction during the Arequipa Mw 8.4, June 23, 2001 earthquake, southern coastal Peru

Eng. Geol.

(2005)

K. Goto et al.

Liquefaction as an important source of the AD 2011 Tohoku-oki tsunami deposits at Sendai Plain, Japan

Geology

(2012)

Y. Hashash et al.

Geotechnical Field Reconnaissance: Gorkha (Nepal) Earthquake of April 25, 2015 and Related Shaking Sequence

(2015)

E. Hauksson et al.

preliminary report on the 29 July 2008 mw 5.4 chino Hills, eastern Los angeles basin, california, earthquake sequence

Seismological Research Letters

(2008)

T.L. Holzer et al.

Liquefaction at Oceano, California, during the 2003 San Simeon earthquake

Bull. Seismol. Soc. Am.

(2005)

Cited by (25)

Geospatial liquefaction probability models based on sand boils occurred during the 2017 M5.5 Pohang, South Korea, earthquake
2024, Engineering Geology
The 2017 Pohang earthquake, triggered by an enhanced geothermal system, had a moment magnitude (M) of 5.5 and caused liquefaction in Pohang City, South Korea. Notably, approximately 600 sand boils were observed only in the soft Quaternary alluvial sediments. This contrasts with most historic liquefaction cases that occurred in earthquakes with M ≥ 6.0. In addition, ground and building settlements were reported in Pohang City. To assess the liquefaction hazard for the Pohang earthquake within the city, we define a 14.4 km² area where the majority of liquefaction was observed. Using the liquefaction data, we develop eight geospatial liquefaction probability models using logistic regression (LR) and random forest (RF) algorithms (five LR- and three RF-based models). These models consider geospatial data (peak ground acceleration, peak ground velocity, slope-derived average shear wave velocity of the upper 30 m, compound topographic index, distance to the nearest water, and roughness) and geotechnical data (averaged standard penetration test N values up to 20 m depth, depth to rock). We evaluate the performances of our eight models and existing global and regional geospatial models using four indicators: accuracy, balanced accuracy, F1 score, and area under the receiver operating characteristic curve. The three RF-based models outperform the other models, followed by the best LR-based model and the global geospatial model from Zhu et al. (2015). Finally, we apply the best RF-based model to the entire Pohang City and evaluate its applicability to other regions and earthquakes. The estimated liquefaction probabilities agree well with the observed distributions of the sand boils and settlements. These geospatial liquefaction models provide a valuable tool for continuously mapping liquefaction hazards in areas with similar geological and geotechnical conditions to Pohang.
Soil moisture-based global liquefaction model (SMGLM) using soil moisture active passive (SMAP) satellite data
2024, Soil Dynamics and Earthquake Engineering
The role of soil saturation condition on the liquefaction occurrence highlights the need for a tool to track the ground-truth soil moisture content involved in this seismic phenomenon. Soil Moisture Active Passive (SMAP) satellite estimates near-real time surface and root zone soil moisture measurements with global coverage. Typical proxies for soil saturation in liquefaction analysis include average water table depth patterns, mean annual precipitation measurements, and topographic conditions. As an alternative to these proxies, this paper incorporates satellite-based soil moisture data to enhance the understanding of the interrelation between saturation conditions and liquefaction events. Proposing a new approach for sampling non-liquefaction cases, a liquefaction/non-liquefaction database was developed in this paper based on reconnaissance reports of eleven target earthquakes. Well-known geospatial explanatory variables as well as new SMAP-based soil moisture parameters, affecting soil liquefaction, are used to develop a new soil moisture-based global liquefaction model (SMGLM), which is compared with an existing global liquefaction model. Considering the ongoing advancement of earth observing satellites, the results of this paper can build a basis for developing fully satellite-based models that could identify liquefied sites using high resolution near-real time soil moisture data.
An uncertainty quantification framework for logistic regression based geospatial natural hazard modeling
2023, Engineering Geology
There is a class of data-driven global natural hazard predictive models that take advantage of broadly available geospatial proxies. These data-driven geospatial models have been commonly used for landslides and are becoming more available in recent years for liquefaction. Logistic regression is the most common method for predicting these ground failure occurrences. These models do not often include robust quantification of uncertainties although they are widely used in the pre-disaster planning and post-disaster response around the world. Taking the logistic regression based global geospatial liquefaction model (GGLM) (Zhu et al., 2017) as an example, we propose an uncertainty quantification (UQ) framework that consists of characterization of different sources of uncertainty, model sensitivity analysis, and forward uncertainty propagation. In this study, we have identified the main sources of uncertainty in such predictive models as parameter estimation uncertainty, modeling error, and geospatial input uncertainty. A Bayesian inference algorithm is used to quantify the posterior distribution of model parameters and quantify model parameter estimation uncertainties which are found to be negligible when a large amount of data is used in the parameter estimation process. Modeling errors are characterized based on the observed residuals between model predictions and measurements and by fitting a Gaussian distribution to the liquefaction probability residuals. The geospatial input uncertainties are characterized using the literature and expert judgment and propagated into model output. Second, we investigate the sensitivity of model output to different uncertain inputs and find that the variance of model output is largely controlled by the geospatial input uncertainties and model errors. Last, we propose an approximate forward uncertainty propagation method, which provides comparable results to a Monte Carlo simulation-based method with better computational efficiency. The proposed UQ framework provides a measure of uncertainty on model predictions and can be applied to any logistic-regression models and other geospatial modeling problems.
Regional seismic risk and resilience assessment: Methodological development, applicability, and future research needs – An earthquake engineering perspective
2023, Reliability Engineering and System Safety
Given the devastating losses incurred by past major earthquake events together with the ever-increasing global seismic exposures due to population growth and urbanization, strategic decision-support tools are required to help stakeholders make more contemplated decisions that promote seismic resilience of the built environment. Such decision-support is enabled by regional seismic risk and resilience assessment, which holistically incorporates the various underlying physics processes and uncertainties for quantitative and probabilistic assessment of regional seismic hazard impacts, and has thereby attracted increasing research focus over the past twenty years. As a significant departure from the traditional site-specific assessment, where only an individual structure and site-specific seismic hazard are of interest, such regional-level assessment introduces additional dimensions and complexity. To date, there is a lack of review studies summarizing the related research advancements in seismic risk and resilience assessment from a regional-level perspective in the context of earthquake engineering. This study fills this gap by conducting a systematic review covering: the methodological development of regional seismic risk assessment (RSRiA) across its key modules, including hazard analysis, exposure modeling, fragility assessment, and consequence evaluation, as well as the associated uncertainty quantification and propagation; the development of resilience metrics, restoration modeling and planning in regional seismic resilience assessment (RSReA); and the applicability of existing computational workflows. Insights into the features, applicability, compatibility, and limitations of existing models and tools are provided. This study also highlights the challenges and future directions toward further advancing the research frontiers of RSRiA and RSReA.
An AI driven, mechanistically grounded geospatial liquefaction model for rapid response and scenario planning
2022, Soil Dynamics and Earthquake Engineering
Citation Excerpt :
Additional observations from Figs. 7–10 and Table 4 are as follows. First, with respect to model bias, the RB20 model was either originally trained (by [5]) or later calibrated (by [6]) using the same observational data adopted herein for testing, except for the Ridgecrest earthquake data which postdated Rashidian and Baise [6]. In contrast, the ML/DL models were trained on CPTs from areas affected by the Loma Prieta, Northridge, and San Simeon events, but not directly on the field observations adopted for testing.
Geospatial models for predicting soil liquefaction infer subsurface traits via satellite remote sensing and mapped information, rather than directly measure them with subsurface tests. Field tests of such models have demonstrated both promising potential and severe shortcomings. Informed by these tests, this paper develops geospatial models that are driven by algorithmic learning but pinned to a physical framework, thereby benefiting both from machine and deep learning, or ML/DL, and the knowledge of liquefaction mechanics developed over the last 50 years. With this approach, subsurface cone penetration test (CPT) measurements are predicted remotely within the framing of a popular CPT model for predicting ground failure. This has three potential advantages: (i) a mechanistic underpinning; (ii) a significantly larger training set, with the model principally trained on in-situ test data, rather than on ground failures; and (iii) insights from ML/DL, with greater potential for geospatial data to be exploited. While liquefaction is a phenomenon best predicted by mechanics, subsurface traits lack theoretical links to above-ground parameters, but correlate in complex, interconnected ways - a prime problem for ML/DL. Preliminary models are trained using ML/DL and a modest U.S. dataset of CPTs to predict liquefaction-potential-index values via 12 geospatial variables. The models are tested on recent earthquakes and are shown – to a statistically significant degree – to perform as well as, or better than, the current leading geospatial model. The models are coded in free, simple-to-use Windows software. The only input is a ground-motion raster, downloadable minutes after an earthquake or available for countless future scenarios. Ultimately, the proposed approach and models, which warrant further application and evaluation, could be improved upon using additional training data and new predictor variables. Users of the models should understand key limitations, as discussed in detail herein.
Compilation and forecasting of paleoliquefaction evidence for the strength of ground motions in the U.S. Pacific Northwest
2021, Engineering Geology
In the U.S. Pacific Northwest (PNW), the historic earthquake record is often insufficient to provide inputs to seismic-hazard analyses or to inform ground-motion predictions for certain seismic sources (e.g., the Cascadia Subduction Zone, CSZ). As a result, paleoseismic studies are commonly used to infer information about the seismic hazard. However, among the many forms of coseismic evidence, soil liquefaction provides the best, if not only, evidence from which the intensities of previous ground motions may be constrained. Accordingly, the overarching goal of this research is to use paleoliquefaction to elucidate previous ground motions in the PNW – both for CSZ events and others – and to further constrain the locations, magnitudes, and recurrence rates of such ruptures. Towards that goal, this paper: (i) reviews current paleoliquefaction inverse-analysis methods and their limited, prior applications in the PNW; (ii) compiles all PNW paleoliquefaction evidence from the literature into a GIS database, resulting in data from 185 study sites (e.g., feature locations, types, sizes, and ages); and (iii) develops maps – specific to the CSZ – that forecast paleoliquefaction for 30 different simulations of a CSZ event. These maps can be used to guide field explorations for new evidence, such that they are conducted efficiently and strategically, considering the apparent utility of evidence towards constraint of CSZ ground-motion models. Of additional utility, this process provides regional ground-motion predictions for physics-based simulations of an M9 event, to include expected site effects. Collectively, the maps of expected shaking intensity and liquefaction may be useful in downstream hazard modeling, regional loss estimation, policy development, and science communication. Ultimately, as more paleoliquefaction evidence is identified and studied, better constraint of regional ground-motion hazards will result.

View all citing articles on Scopus

View full text

Regional efficacy of a global geospatial liquefaction model

Highlights

Abstract

Introduction

Section snippets

Database and methodology

Development of a GGLM-based liquefaction intensity index

Conclusion and key remarks

Declaration of Competing Interest

Acknowledgement

Soil Dyn. Earthq. Eng.

Eng. Geol.

Soils Found.

Soils Found.

Eng. Geol.

Soils Found.

Quat. Int.

Eng. Geol.

Eng. Geol.

Eng. Geol.

Eng. Geol.

Eng. Geol.

Soil Dyn. Earthq. Eng.

Soils Found.

Eng. Geol.

Initial Geotechnical Observations of the Bhuj, India, Earthquake of January 26, 2001

Some Observations of the Geotechnical Aspects of the February 28, 2001, Nisqually Earthquake in Olympia, South Seattle, and Tacoma, Washington

Geotechnical Engineering Reconnaissance of the August 24, 2014 M6 South Napa Earthquake

Geotechnical Reconnaissance of the 2015 Illapel Earthquake

Geotechnical Quick Report on the Affected Region of the 23 August 2011 M5. 8 Central Virginia Earthquake near Mineral, Virginia

Ground Failure Associated with the Puget Sound Region Earthquakes of April 13, 1949, and April 29, 1965 (No. 90–687).

The Geotechnical Aspects of the September 3, 2016 M 5.8 Pawnee, Oklahoma Earthquake

Geotechnical Reconnaissance of the 2010 Darfield (New Zealand) Earthquake

Geotechnical Reconnaissance of the 2011 Christchurch, New Zealand Earthquake

23 October 2011 Van (Turkey) earthquake

Nat. Hazards

Global patterns of groundwater table depth

Science

Soil liquefaction during the Arequipa Mw 8.4, June 23, 2001 earthquake, southern coastal Peru

Eng. Geol.

Liquefaction as an important source of the AD 2011 Tohoku-oki tsunami deposits at Sendai Plain, Japan

Geology

Geotechnical Field Reconnaissance: Gorkha (Nepal) Earthquake of April 25, 2015 and Related Shaking Sequence

preliminary report on the 29 July 2008 mw 5.4 chino Hills, eastern Los angeles basin, california, earthquake sequence

Seismological Research Letters

Liquefaction at Oceano, California, during the 2003 San Simeon earthquake

Bull. Seismol. Soc. Am.