A rainfall runoff model based on a digital elevation model (DEM) was applied to a small catchment in Happy Valley, South Australia to predict catchment storm runoff. The DEM was used to partition the catchment into several thousand irregular shaped elements. These elements, with an average size of 825 m2 each, form an interconnected one-dimensional flow network for runoff routing. The rainfall runoff model is a kinematic flow model which combines the solving of flow continuity equation and the Manning's equation to generate surface and subsurface runoff. This study improves on the existing rainfall runoff model in several areas. It adds spatial rainfall averaging methods to derive spatial rainfalls for catchment modelling; and it improves the catchment soil moisture representation by developing a boundary wetness index, and relates this index to antecedent catchment flow to derive spatial catchment moisture distribution. Improved runoff predictions were obtained as a result of the improvement in spatial data input and spatial soil moisture representation. The study identifies these improvements as the key areas for better runoff prediction. It demonstrates that where prediction results showed larger than expected variance, it is frequently caused by the inability to derive good spatially distributed input data rather than parameter estimation errors.