中文摘要：

针对GRACE时变重力场球谐系数的高阶次项中含有的较高噪声会造成反潢结果产生误差的问题,该文采用多通道奇异谱分析和高斯滤波相结合的方案来滤除GRACE数据中的噪声,并与普遍使用的去相关滤波和高斯滤波相结合的方案进行了对比验证。首先利用这两种方法分别计算了2007年4、10月份的全球水储量变化,结果显示,这两种方法反演的结果基本相同,而多通道奇异谱分析与高斯滤波相组合的方法提高了信噪比,减弱了信号泄露误差,验证了该组合方案的可行性。然后又分别计算了亚马逊流域2004—2010年的水储量时间变化序列,并与GLDAS水文模型进行了对比验证,结果显示,这两种方法得到的水储量时间序列变化趋势基本相同,各自之间具有强相关性,进一步说明了该文方案在GRACE数据滤波中的可行性。

英文摘要：

High-degree and high-order spherical harmonics of time-variable gravity fields observed by GRACE are dominated by noise.In this article,the noise existing in GRACE data was filtered by the combination of multi-channel singular spectrum analysis（MSSA）and Gaussian filter.The global terrestrial water storage changes（TWSCs）in April and October 2007 were calculated by MSSA and empirical de-correlation method（EDM）in combination with Gaussian smoothing respectively,and the almost same results were obtained by these two methods;meanwhile,the proposed filter method could improve the ratio of signal-to-noise,lower signal leakage errors more effectively,so it could be well applied in deducing the TWSCs.The TWSCs of Amazon basin was also estimated by GRACE from 2004 to 2010,and the results of GLDAS on temporal distribution showed a strong correlation with the correlation coefficient,which verified the feasibility of the proposed filter method.