Abstract: To decrease computation time and consequently improve computation efficiency of GPM, parallel algorithms based on pollution sources, levels of study area and grids of each level were designed. These parallel algorithms were also implemented and run on computer cluster. Our efforts were tested with a case of air pollution dispersion in Pearl River Delta. The test result indicates that the parallel algorithms can decrease computation time vastly by ninety percent, which makes it is possible to apply GPM in emergence response modeling and computation.