Stochastic Risk Analysis of Settlement in Large‐Filled Areas Considering the Local Spatial Effect of Soil Properties

ABSTRACT Reliable prediction of final settlement and practical risk assessment are essential for the economic design of large‐scale land reclamation projects. While stochastic analysis offers valuable insights into soil spatial variability, the critical issue of large‐scale statistical heterogeneity has been overlooked in most previous work. Given that reclamation is typically conducted over extensive areas, the statistical parameters of soil properties are unlikely to be uniform across the entire site. This study highlights these local spatial effects and presents a novel stochastic analysis approach to account for them in final settlement and risk evaluations for a reclaimed area. Clustering analysis reveals two distinct zone types within the study area: a region exhibiting well‐clustered borehole data and another characterized by dispersed, non‐clusterable data. For the former, considering statistical parameters from the clustered area in the stochastic analysis yields a more economical and safer risk assessment. For the latter, a safe design is ensured by adopting the parameter set, whether global or from an overlapping zone, with the highest coefficient of variation in the compression index. The findings indicate that overlooking these critical local effects by using only global statistical parameters can result in a design that is either less safe or unnecessarily costly.