Advisor: Byungmin Kim Department of Urban and Environmental Engineering Ulsan National Institute of Science and Technology (UNIST). Dissertation committee member: Young Joo Lee Department of Urban and Environmental Engineering Ulsan National Institute of Science and Technology (UNIST). Dissertation Committee Member: Han-saem Kim Earthquake Research Center Korea Institute of Geoscience and Mineral Resources (KIGAM).
Unlike the 2016 M5.5 Gyeongju earthquake in South Korea, the strongest instrumented earthquake, the Pohang earthquake occurred in an area of Quaternary sediments with thick fill and alluvial layers, causing more damage due to ground motion amplification and liquefaction. Among other ground deformations such as subsidence, cracks and landslides, the main problem was the hundreds of liquefaction sand boils observed near the epicenter. This study collected approximately 2000 Standard Penetration Test (SPT) N values from the Geotechnical Information Portal, National Disaster Management Research Institute, Korea Meteorological Administration, and local government offices.
This study calculated the Liquefaction Potential Index (LPI) and Liquefaction Severity Number (LSN) using this data. It was found that the higher LPI and LSN values correspond to the locations of sand boiling. In addition, this study calculated the probability using the number of girds where sandboils occurred and no sandboils were observed.
Additionally, this study calculated the geospatial liquefaction probability model using sand boiling locations and influencing factors that reflect ground motion intensity and geological and soil characteristics (i.e., peak ground acceleration (PGA), peak ground velocity (PGV), composite topographic index (CTI) from the digital elevation model (DEM), average SPT-N value for the highest 5 m soil deposits, distance from the river near sand boiling locations, VS30 and depth to rock).
Introduction
LPI and LSN
- Previous studies
- Data
- Methodology
- LPI
- LSN
- Results
The 2017 Pohang earthquake generated hundreds of sand boils mostly in the rice fields as shown in Figure 3(a). The rice fields where most of the sand boils are within a basin and are surrounded by mountains. Figure 3(b) shows the geology (from the Korea Institute of Geoscience and Mineral Resources (KIGAM)) in the vicinity of the study area.
Moreover, the sand boils were concentrated near the epicenter: 96.2 % of the sand boils occurred within a radius of 4 km from the epicenter, and 80.53% within a radius of 2 km. Most of the sand boils occurred in the areas with PGA of 0.2 g as shown in Figure 3(d). For the ranges shown, values approximately central to the middle third of the range are more common than outlying values, but ER and 𝐶𝐸 can be even more highly variable than the ranges shown if equipment and/or monitoring and procedures are not good is not.
LPI values range from 0 to 100 and the degree of liquefaction damage is in accordance with the LPI range shown in the table below. Since CRR is higher at N1.60, CRR was not shown in the panel (Boulanger and Idriss 2014). Most of the sand boiling occurred in the PGA 0.2 g areas, as shown in Figure 3(d) .
LPI values were calculated to be high in the two wells (ie, 16 and 19) located in the center of the basin (within the study area). Therefore, the spatial interpolation resulted in the highest LPI values concentrated in the center of the basin, which are consistent with the spatial distribution of sand boiling observations (Figure 9a). A similar pattern can be observed for LSN: LSN is highest in the center of the basin (i.e., 135) and approaches zero toward the edge of the basin, also consistent with the distribution of sand boiling (Figure 9b).
Unlike the LPI case, some wells at the edge of the basin have non-zero LSN values, resulting in wider areas with non-zero LSN within the basin. About 55% of liquefied networks and 20% of non-liquefied networks have estimated LPIs higher than 5 (true positive or false positive, respectively). About 8% of liquefied networks and 2% of non-liquefied networks have estimated LPIs higher than 15.
This discrepancy between the mean LPIs for this study and previous studies could be due to the different severity of liquefaction considered in both studies: the liquefaction observed in this study is mostly, while those of Iwasaki include all types of severity (that i.e., from marginal to moderate and severe liquefaction). The numbers of sand boil observations not shown in the graph are shown above the panels.
Geospatial Liquefaction Probability model
- Previous studies
- Data
- Methodology
- Results
The input factors and the presence of sand boiling are assigned to grids m) used in the logistic regression. Maps of PGA (peak ground acceleration) and PGV values estimated for the Pohang earthquake were obtained from USGS ShakeMap (https://earthquake.usgs.gov/data/shakemap/) (Figure 15a and Figure 15b). Figure 15(c) shows that the CTI values are higher than 16 in the basin (near the study area), and the mountainous areas have lower CTI values.
Figure 15(e) shows the depths to rock (𝑫𝒓𝒐𝒄𝒌) determined using the collected borehole data mentioned in Section 2.2. 𝑵̅𝟓𝒎, VS30 and 𝑫𝒓𝒐𝒄𝒌, which were obtained from the borehole data, were spatially interpolated using the KRIGING tools in the GIS program. This study loaded the mean values of data into the grid and calculated the liquefaction probability model using the previously described data.
These different models include factors from each of the three categories (soil intensity, soil intensity and water content), as the PGA is indispensable. This study considers the Model 1 as a base model consisting of two variables: PGA and 𝑁̅5𝑚. In addition, this study uses CTI and Rriver to consider the effect of water content.
According to Table 10, this study selected model 10 as the most successful model using PGA, 𝑁̅5𝑚, CTI, Drock and VS30 as variables. The accuracy values for the best model are greater than those of the other two models at all thresholds. This study evaluated the effect of different shear wave velocity averaging depths (i.e. VS5, VS10, VS20 and VS30) for the best model.
This study calculated the receiver operating characteristic (ROC) curve and compared it with the previous study to evaluate the proposed liquefaction probability model. According to Figure 19, the ROC curve of this study is in the upper part of the graph than the results of the previous study, which implies that the model proposed by this study performs slightly better than the previous study. Overall, the probability of the best model ranges from 0 to 0.7, and that of the base model from 0 to 0.4.
The best model gave higher probability in the area where most of the sand boils are concentrated. Therefore, the best model proposed by this study is good to apply in Korea.
Conclusions
34; Empirical Near-Source Attenuation Relations for Horizontal and Vertical Components of Peak Ground Acceleration, Peak Ground Velocity, and Pseudo-Absolute Acceleration Response Spectrum." Seismology Research Papers. )-Turkey Earthquakes." Earth Dynamics and Earthquake Engineering. 34; An Update of the Simplified Seed-Idriss Procedure for Estimation of Liquefaction Potential." Proc., TRB Workshop on New Approaches to Liquefaction, Pubbl.
A practical method for evaluating soil liquefaction potential based on case studies at various locations in Japan. 34;Geospatial post-earthquake hazard assessment of the 2017 Pohang earthquake considering seismic site effects." ISPRS International Journal of Geo-Information 7(9): 375. 34;Evaluation of liquefaction potential index for hazard assessment Liquefaction in Christchurch, New Zealand". Journal of Geotechnical and Geoenvironmental Engineering.
34;Investigating the discrepancy between forecast and observed liquefaction from the 2015 Gorkha, Nepal, earthquake.” Earthquake Spectra 33(S1): S73-S83.
