A quadtree approach to domain decomposition for spatial interpolation in Grid computing environments

Spatial interpolation is widely used in geographical information systems to create continuous surfaces from discrete data points. The creation of such surfaces, however, can involve considerable computation, especially when large problems are addressed, because of the need to search for neighbors on which to base interpolation calculations. Computational Grids provide the computing resources to tackle spatial interpolation in a timely way. The objective of this paper is to investigate the use of domain decomposition for a distributed inverse-distance-weighted spatial interpolation algorithm; the algorithm runs using the Globus Toolkit (GT) in a heterogeneous Grid computing environment. The interpolation algorithm is modified for implementation in the Grid by using a quadtree to spatially index and adaptively decompose the interpolation problem to balance processing loads. In addition, the GT allows the distributed algorithm to couple multiple machines, potentially of different architectures, to dynamically schedule the decomposed sub-problems through Globus services and protocols (e.g., resource management, data transfer). Experiments are conducted to test how well this distributed IDW interpolation algorithm scales to heterogeneous grid computing environments using irregularly distributed geographical data sets.