Application Of Parallel Computing To Gravity Field Recovery FromSatellite Gravity Gradiometric Data
Application Of Parallel Computing To Gravity Field Recovery FromSatellite Gravity Gradiometric Data
复制标题
并行计算在卫星重力梯度数据重力场恢复中的应用
DOI：
10.2495/hpc000101
复制
发表时间：
2000
期刊：
影响因子：
--
通讯作者：
and P.N.A.M. Visser
中科院分区：
文献类型：
--
作者：
R. V. Geemert;R. Koop;R. Klees;and P.N.A.M. Visser
重力场和稳态海洋环流探测器（GOCE）是欧洲航天局（ESA）为“地球探测器”计划实施而进行的一项重力场使命。该使命的目标是提供一个独特的地球重力场全球模型，具有前所未有的准确性和空间分辨率。这将为固体地球物理学、海洋学和大地测量学等广泛的学科提供新的和基本的见解。将通过结合利用星载重力梯度仪进行的重力梯度测量和利用全球定位系统/全球轨道导航卫星系统接收器进行的卫星间测量来实现使命目标。从数值的角度来看，从这些观测值估计重力场参数是一个真实的挑战。大约62500重力场参数估计在最小二乘法从超过100万的观测收集在12个月的使命寿命。将观测值与重力场参数相关联的线性模型不允许显式地计算和存储正规方程。因此，现有的线性代数软件包的MIMD机是没有什么用的MIMD机和特殊的算法必须设计。我们将讨论在德尔夫特理工大学的CRAY-T3 E超级计算机上开发GOCE的SGG和SST数据的并行数据处理设备。初步结果表明，该算法的问题特定的设计允许估计重力场参数从GOCE SGG/SST数据。我们讨论了我们的算法的并行度，可扩展性和所需的等待时间方面的性能。
The Gravity field and steady-state Ocean Circulation Explorer (GOCE) is a ded-icated gravity field mission currently under assessment by ESA (European Space Agency) for implementation in the 'Earth Explorer' Programme. The objective of the mission is to provide a unique global model of the Earth's gravity field with un-precedented accuracy and spatial resolution. This will provide new and fundamen-tal insight into a wide range of disciplines such as Solid Earth Physics, Oceanog-raphy and Geodesy. The mission goals will be achieved from a combination of gravity gradient measurements by a spaceborne gradiometer (SGG) and satellite-to-satellite (SST) measurements by a GPS/GLONASS receiver. The estimation of the gravity field parameters from these observations is a real challenge from a numerical point of view. About 62500 gravity field parameters have to be estimated in a least-squares approach from more than 100 million observations collected during the mission lifetime of 12 months. The linear model relating the observations to the gravity field parameters does not allow calculating and storing the normal equations explicitly. Therefore, the available linear algebra software packages for MIMD machines are of little use and special algorithms for MIMD machines have to be designed. We will discuss the development of a parallelized data processing facility for SGG and SST data from GOCE on the CRAY-T3E supercomputer at Delft University of Technology. First results indicate that a problem-specific de-sign of the algorithms allows estimating the gravity field parameters from GOCE SGG/SST data. We discuss the performance of our algorithm in terms of degree of parallelism, scalability properties and required waiting times.