The spatial variability of rainfall at the soil surface and tree roots in a coastal sand dune forest was investigated. Various scenario simulations were further conducted using calibrated HYDRUS models to assess the effects of these vegetation-related variability patterns on soil moisture dynamics and the water balance. Vadose zone soil moisture and groundwater recharge in forested ecosystems can be influenced by canopy architecture and root systems. Spatial distributions of surface rainfall and tree roots were observed at the canopy and intercanopy areas in a coastal sand dune forest of subtropical Australia. We further explored the effects of these vegetation-associated spatial patterns on the spatiotemporal dynamics of soil moisture and deep drainage in this system using calibrated HYDRUS models. Simulated soil moisture and deep drainage were found higher at the intercanopy area than the canopy area due to the reinforced effects of rainfall interception and root water uptake. Spatial analysis of deep drainage at various locations across the canopy–intercanopy transect resulted in an average deep drainage of 1347 mm yr−1 (65% of annual gross rainfall), with 68% of total deep drainage occurring at the intercanopy area. The scenario analyses indicate that the spatial variability of rainfall at the soil surface is not important for the larger scale groundwater balance in our coastal sand dune forest. However, deep drainage was underestimated by 130 to 162 mm (9.7–12.0% drop compared with baseline scenario) as a consequence of uniform representation of heterogeneous root systems in one- or axisymmetric three-dimensional HYDRUS models. Our HYDRUS simulations indicated that translating the hydrological effects of the two-dimensional tree structure (rainfall redistribution and heterogeneous roots) to a one-dimensional lumped vertical conceptualization in coastal sand dune forests needs to be undertaken with caution.