Regional efficacy of a global geospatial liquefaction model
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 (LIIGGLM) 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)
- et al.
Liquefaction of soil in the Tokyo Bay area from the 2011 Tohoku (Japan) earthquake
Soil Dyn. Earthq. Eng.
(2011) - et al.
Toward a numerical-simulation-based liquefaction hazard assessment for urban regions using high-performance computing
Eng. Geol.
(2019) - et al.
Liquefaction-induced ground displacement and its related damage to lifeline facilities
Soils Found.
(1996) - et al.
Case studies of liquefaction in the 1964 Niigata earthquake
Soils Found.
(1981) - et al.
Liquefaction potential index considering parameter uncertainties
Eng. Geol.
(2009) - et al.
Reconnaissance report on geotechnical damage caused by an earthquake with JMA seismic intensity 7 twice in 28h, Kumamoto, Japan
Soils Found.
(2016) - et al.
A lower Pleistocene to Holocene terrestrial record from the Eskişehir Graben (Central Anatolia): Paleoclimatic and morphotectonic implications
Quat. Int.
(2019) - et al.
Assessment of liquefaction potential for two liquefaction prone areas considering the May 20, 2012 Emilia (Italy) earthquake
Eng. Geol.
(2015) - et al.
Recurrent liquefaction-induced failures triggered by 2014 Cephalonia, Greece earthquakes: spatial distribution and quantitative analysis of liquefaction potential
Eng. Geol.
(2016) - et al.
A kriging method of interpolation used to map liquefaction potential over alluvial ground
Eng. Geol.
(2013)
Liquefaction hazard mapping by liquefaction potential index for Dhaka City, Bangladesh
Eng. Geol.
Modification of the liquefaction potential index to consider the topography in Christchurch, New Zealand
Eng. Geol.
Coastal failures during the 1999 Kocaeli earthquake in Turkey
Soil Dyn. Earthq. Eng.
Liquefaction of the ground during the 1978 Miyagiken-Oki earthquake
Soils Found.
Liquefaction-induced ground failure: a study of the Chi-Chi earthquake cases
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.
Cited by (25)
Soil moisture-based global liquefaction model (SMGLM) using soil moisture active passive (SMAP) satellite data
2024, Soil Dynamics and Earthquake EngineeringRegional seismic risk and resilience assessment: Methodological development, applicability, and future research needs – An earthquake engineering perspective
2023, Reliability Engineering and System SafetyAn AI driven, mechanistically grounded geospatial liquefaction model for rapid response and scenario planning
2022, Soil Dynamics and Earthquake EngineeringCitation 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.