ESTIMATION OF HYDRAULIC CONDUCTIVITY AND SWCC PARAMETERS FOR UNSATURATED SOIL BASED ON FIELD MINI DISK INFILTROMETER TESTS AND MERGING PARTICLE FILTER

Data assimilation methods, such as the Merging Particle Filter (MPF), have demonstrated effectiveness in estimating soil properties within geotechnical engineering. However, due to the significant variability of hydraulic conductivity across soil types, the accurate estimation of this parameter remains challenging. Moreover, previous assimilation approaches assume uniformly prepared soils in laboratory settings, limiting their ability to evaluate the pronounced spatial variation typically observed in field soils. To overcome these limitations, this study proposes a sequential data assimilation framework that integrates field mini disk infiltrometer tests with MPF for rapid evaluation of field hydraulic characteristics. The resulting posterior distributions of soil parameters successfully reproduce field conditions, validating the effectiveness of the proposed framework. Additionally, the effects of model dimensionality, particle number, and observation noise level on assimilation accuracy are systematically examined. Compared to the two-dimensional (2D) model, the three-dimensional (3D) model yields more accurate and robust results. Furthermore, careful selection of particle number and observation noise level is crucial, as excessively high particle numbers and unrealistically low noise levels do not necessarily enhance model performance or reliability.