Mapping Approaches for Assessing Regional Soil Liquefaction Potential: A Comparative Assessment
Abstract
Seismic liquefaction poses a significant threat to infrastructure and human safety, making accurate regional-scale hazard assessment essential for effective disaster mitigation. This study systematically compares simple kriging (SK), ordinary kriging (OK), and sequential Gaussian simulation (SGS) for liquefaction potential mapping using a case study of the Chi-Chi earthquake. Results show that among the eight theoretical semivariogram models, the exponential model was identified as the optimal semivariogram model, showing the best fit and the lowest validation errors. SK and OK yield nearly identical LPI estimates with high accuracy and low computational cost, but oversmooth high-risk zones and underestimate spatial uncertainty. In contrast, SGS captures spatial heterogeneity and extreme values, with variance and coefficient of variation maps revealing higher and more heterogeneous uncertainties, which offers a more comprehensive basis for probabilistic risk assessment. However, these advantages came at a substantially higher computational cost. The choice of method should therefore depend on the engineering objective: kriging for rapid, large-scale trend estimation, and SGS for detailed probabilistic risk evaluation where capturing extreme values and uncertainty is critical.