Design of parallel algorithm for spatial sensitivity analysis based on MapReduce

Design of parallel algorithm for spatial sensitivity analysis based on MapReduce

Li Fan He Honglin Jinghinli Zhang Lu Ying Yuguiri

In recent years, with the wide application of remote sensing spatial data in ecosystem, the development of regional scale ecological Remote sensing parameter model has been promoted. Sensitivity analysis plays an important role in identifying key parameters of model, reducing model uncertainty and perfecting model. In the sensitivity analysis of the model parameters, the stand-alone environment can not meet the requirement of fast analysis because of the complex operation of spatial data. In order to improve the efficiency of spatial sensitivity analysis of ecological Remote sensing parameter model, this paper designs and realizes the Vegetation based on the Qinghai-Tibet Plateau as the research area, using the vegetation photosynthetic model VPM (photosynthesis models) and the open source cloud computing platform Hadoop. The spatial sensitivity analysis of the ecological Remote sensing parameter model, and the algorithm is analyzed in the laboratory cluster environment, which verifies the validity and applicability of the algorithm. The core of the algorithm is to use the MapReduce parallel programming technology to segment the map sampling and the iterative process of the model in the spatial sensitivity analysis, and to assign the segmented subtasks to different compute nodes for parallel computation. The experiment shows that the parallel algorithm proposed in this paper can effectively shorten the time of map sampling and model iterative calculation, and the running speed of parallel algorithms is about 14 times times higher than that of single machine algorithm.

Design of parallel algorithm for spatial sensitivity analysis based on MapReduce