High-Performance GeoComputation with the Parallel Raster Processing Library

The recent developments in GIScience and GeoComputation have seen the emergence of big spatial data and complex geospatial algorithms, which greatly increase the demand for high-performance computing. However, developing an effective and efficient parallel GeoComputational algorithm requires taking into account the unique characteristics of the spatial data and algorithm, leading to extensive development complexity. The parallel Raster Processing Library (pRPL) is an open-source programming library designed for GIScientists and GeoComputation practitioners to implement parallel raster-based geospatial algorithms. While providing easy-to-use interfaces for users, pRPL automatically takes care of the underlying parallel computing procedures (including domain decomposition, assignment mapping, algorithm execution, data exchange, load-balancing, and data I/O), therefore largely reducing the development complexity. After giving a brief introduction to pRPL 2.0, this paper presents two showcases of using pRPL to implement high-performance spatial computing: slope/aspect calculation and Cellular Automata modeling. The experiments show that high-performance GeoComputation could be implemented using pRPL by those with minimal parallel programming skills.