Modelling of unstable fingered flow in unsaturated soil with Gaussian random fields




Modelling of unstable fingered flow in unsaturated soil with Gaussian random fields


Hydrological processes in soils, such as fingering and preferential flow, are strongly related to the heterogenous nature of the medium. Such processes can result in accelerated transport of solutes, whilst also causing localised erosion. Neglecting the distinct variation in material properties in numerical simulations can result in an unrealistically uniform wetting front. Here, we represent the variability of the soil mass by random fields, generated through the solution of Whittle-Matérn priors. By solving a stochastic PDE, we avoid the need for large covariance matrices often present in random field generation. As the generated field is an arbitrary representation of correlated space, it can be assigned to represent any spatially varying property. Similarly, it is possible to generate multiple fields for representing spatially varying parameters whose correlation structures differ. Here, we choose the saturated hydraulic conductivity and van Genuchten parameters as they directly affect the hydraulic response of moisture transfer. In doing so, we find that there is accelerated transport in more conductive areas and illustrate this with an example for a hydrophobic soil.



Evan Ricketts; P. Cleall; A. D. Jefferson; Pierre Kerfriden; Paul Lyons


10th European Conference on Numerical Methods in Geotechnical Engineering (NUMGE2023)



2. Finite element, finite difference, discrete element, material point and other methods



https://doi.org/10.53243/NUMGE2023-304