Large scale distributed hydrologic modelling using a massively parallelised two dimensional finite volume hydrodynamic model

The paper introduces a computationally fast two dimensional hydrodynamic model applicable to hydrologic and flood runoff modelling assessment. The model, TUFLOW GPU, solves the shallow water equations using a parallelised, explicit, 1st order finite volume numerical scheme compiled for computation via GPU processors rather than CPU processors. This numerical approach allows significant increases in computational performance; the paper demonstrates benchmark tests of onehundred- fold increases in computational speed over TUFLOW Classic. The approach is ideal for application to large scale simulation of direct rainfall inundation and subsequent infiltration and runoff, such as for inland Australia where vast tracts of land can become inundated and flow paths are often difficult to pre-define. The paper presents an application to the Condamine-Balonne Basin, with a model area of 533,000km2, and simulates direct rainfall, infiltration losses via the Green-Ampt method and subsequent inundation of land. This application uses a 180 million cell model with a grid resolution of 62m over the study area. The paper discusses the benefits of this approach over other more conventional approaches (such as lumped hydrologic modelling) and highlights key issues associated with its successful implementation.