Recharge assessment in fractured (karst) aquifers is an essential component in groundwater management and in vulnerability estimation, concerning both, the quantification of the delay of the arrival of the recharge pulse at the groundwater table and of the storage volume of the vadose zone. In contrast to diffuse infiltration, frequently encountered in consolidated and unconsolidated porous media, the infiltration dynamics in the unsaturated zone of fractured-porous rocks and karst aquifers exhibits a rapid, gravity-driven flow component along preferential flow paths such as fractures, fracture networks, faults and fault zones. The partitioning into two hydraulically contrasting domains commonly leads to a breakdown of classical volume-effective flow equations employed in many FD or FEM modeling approaches which only consider the capillarity of the porous medium. Even in the presence of a porous matrix, preferential pathways along fractures have been shown to sustain flow percolation for equilibrium and non-equilibrium conditions. In order to properly capture the gravity driven film flow or drop flow physics along preferential vertical flow paths, various factors have to be considered such as static and dynamic contact angles, surface tension, free-surface (multi-phase) interface dynamics, dynamic switching of flow modes (between droplets, rivulets, films) and associated formation of singularities in the case of merging or snapping flow. Specifically, this presentation addresses (1) limitations of unsaturated multi-continuum catchment-scale simulations, (2) process-oriented infiltration dynamics along preferential flow paths with a parallelized 2D/3D smoothed particle hydrodynamics approach (SPH), and (3) pore-scale multiscale SPH modeling approaches to handle fracture-matrix interactions and transport phenomena in the small-scale limit when the advection-diffusion equation breaks down. Laboratory experiments were carried out to study some of the phenomena encountered, validate our code and obtain analytical transfer functions to study the given system in the large-scale limit.