海域重力场解算中垂线偏差方法的并行化改进

Parallelization improvement of vertical deviation method in the calculation of sea gravity field

垂线偏差法是卫星测高数据反演海洋重力异常的主要途径之一.然而全球卫星测高数据的积分计算量庞大,传统的串行计算需要耗费大量时间.为了提高垂线偏差法反演海洋重力异常的计算效率,本文以逆Vening-meinesz公式为基础,提出了基于GPU并行的垂线偏差反演方法.该方法利用GPU强大的并行计算能力,将垂线偏差反演中核函数、方位角等数量庞大的计算分割成了成百上千的独立单元,相对于串行计算极大地提高了计算效率.文章利用EGM2008模型大地水准面高进行测试,并将改进后的GPU并行方法与经典的垂线偏差反演方法进行对比,结果表明采用GPU并行的垂线偏差反演方法在保证计算精度的情况下,可以大幅提高计算速度,测试中加速比最大可达到23.3,对于利用卫星测高数据快速解算高分辨率海洋重力场有重要的意义.

The vertical deviation method is one of the main ways to retrieve marine gravity anomaly from satellite altimetry data. However, the integral calculation of global satellite altimetry data is huge, the traditional serial calculation requires a lot of time. In order to improve the computation efficiency of inversion of marine gravity anomaly by vertical deviation, a vertical deviation method based on GPU parallel is proposed in this paper, which is based on inverse Vening-Meinesz formula. This method make full use of powerful parallel computing capability of GPU and divide the computation of kernel function and azimuth angle in vertical deviation inversion into hundreds and thousands of independent units, which greatly improves the computational efficiency compared with serial computation. In this paper, the geoid height of EGM2008 is used to test, and the improved method is compared with the classical vertical deviation inversion method. The results show that the GPU parallel vertical deviation inversion method can greatly improve the calculation speed, and the maximum acceleration ratio can reach 23.3, while the calculation accuracy remains unchanged. It is of great significance to use satellite altimetry data to quickly calculate the high resolution ocean gravity field.