A full graphics processing unit implementation of uncertainty-aware drainage basin delineation

Terrain analysis based on modern, high-resolution Digital Elevation Models (DEMs) has become quite time consuming because of the large amounts of data involved. Additionally, when the propagation of uncertainties during the analysis process is investigated using the Monte Carlo method, the run time of the algorithm can increase by a factor of between 100 and 1000, depending on the desired accuracy of the result. This increase in run time constitutes a large barrier when we expect the use of uncertainty-aware terrain analysis become more general. In this paper, we evaluate the use of Graphics Processing Units (GPUs) in uncertainty-aware drainage basin delineation. All computations are run on a GPU, including the creation of the realization of a stationary DEM uncertainty model, stream burning, pit filling, flow direction calculation, and the actual delineation of the drainage basins. On average, our GPU version is approximately 11 times faster than a sequential, one-core CPU version performing the same task.