Seepage analysis model based on field measurement data for estimation of posterior parameter distribution using merging particle filter

Abstract

Soil water conditions should be adequately evaluated because they influence the occurrence of surface failures. Digital twin systems, connecting field measurement data with numerical simulations, must be created to enable early warnings to be issued before a surface failure occurs. This study discusses the applicability of the merging particle filter (MPF) method for estimating the posterior distribution of seepage analysis models based on the volumetric water content field measurement data from two case studies. The first case study estimated the posterior distribution of parameters for unsaturated soil hydraulic properties based on data obtained from three slopes of different soil types (decomposed granite, weathered mudstone, and pyroclastic flow deposits). The simulation results agreed well with the raw data, where only precipitation data were input into the estimated seepage analysis model. The second case study estimated and discussed the applicability of a seepage analysis model using parameters for the unsaturated soil hydraulic properties and drainage boundary conditions. The simulated results reproduced the field measurement data with sufficient accuracy to attain the groundwater behavior. Therefore, based on field measurement data, the MPF can estimate the posterior distribution of parameters for the seepage analysis model, considering the inhomogeneity and uncertainty of in-situ soil.

Keywords

• Data assimilation
• Merging particle filter
• Unsaturated soil hydraulic properties
• Boundary condition
• Field measurement data