摘要
为了加快气溶胶光学厚度(AOD)反演计算速度,基于SYNTAM串行算法,提出了循环分块划分和聚合通信的策略,利用消息传递模型,在中国气象局的IBM Cluster 1600高性能计算机系统上,并行实现了从MODIS双星(TERRA和AQUA)卫星数据反演AOD。试验结果表明该方法大大减少了计算时间,与地面太阳光度计实测AOD数据进行对比验证,发现所有站点处的AOD反演相对误差小于22%,表明这种并行方法可以满足高精度监测空气质量要求。
To speed up the computation of the retrieval of Aerosol Optical Depth （AOD）, a cyclic partition and collective communication strategy based on serial computation algorithm of Synergy of TERRA/MODIS and AQUA/MODIS （SYNTAM） was proposed to archive parallel computation of AOD retrieval from the both Moderate Resolution Imaging Spectroradiometer （MODIS） satellite （ TERRA, AQUA） data by message passing on the IBM cluster 1600 high performance compute of Chinese Meteorological Administration. Expe.rimentaI results demonstrate that the parallel SYNTAM can decrease computation time greatly. Compared with measurements from ground-based sun-photometers, the relative error of experiment results from the parallel SYNTAM is less than 22% in call cases, indicating that it can meet the needs of high accuracy such as air quality monitoring.
出处
《计算机应用》
CSCD
北大核心
2009年第6期1665-1668,共4页
journal of Computer Applications
基金
国家973计划项目(2007CB714407O7S00502CX)
中央公益性基本科研业务专项项目(2007Y001)
关键词
气溶胶光学厚度
并行计算
循环分块
消息传递接口
Aerosol Optical Depth （AOD）
parallel computation
cyclic partitioning
Message Passing Interface （MPI）
