Parallel geospatial analysis on windows HPC platform

Windows HPC is a high performance computing platform which has been applied in various areas. However, by reviewing the applications list, geospatial analysis is still one of the areas which Windows HPC has not touched a lot, such as Inverse Distance Weighting (IDW), a typical method in geospatial analysis. IDW interpolation assigns values to unknown points by calculating a weighted sum of values from known points. The whole process of IDW consists of domain decomposition, spatial interpolation, output gathering and data visualization. In order to execute IDW on Windows HPC, we modify and recompile IDW source package to port it to HPC environment and configure a multi-task job. By using spatial domain with different number of clustering regions as input data, IDW generates different interpolation output which can be rendered into different images. The experiments test four cases, and show Windows HPC can improve the efficiency of IDW without any loss in accuracy.