high performance computing on graphics processing units: hgpu.org

Most data for environmental variables (e. g. meteorological variables, soil properties etc.) are collected from point sources. For modeling and visualization purposes, the data is often needed to be available on a regular grid, which requires spatial interpolation of the scattered point measurements. A variety of interpolation methods for these purposes is available, examples are inverse distance weighting (IDW), Kriging, splines or polynomial regressions. Depending on the number of prediction locations (i. e., the grid size) as well as the number of data points, interpolation can be a very time- and performance consuming task. Subject of this work was to implement the inverse distance weighting algorithm, which is a very simple interpolation method but widely used in geoscientific applications, on a GPU. The properties of the algorithm make it well suited for performing the calculation on a GPU rather than on the CPU, by doing so expecting a significant performance gain. Similar studies transferring spatial interpolation algorithms onto the GPU [4, 5, 9] show promising results.