基于改进的主成分分析法提取GRACE时变重力场信号

Extraction of GRACE Time-Varying Gravity Field Signal Based on Improved Principal Component Analysis

针对主成分分析法提取时变重力场模型信号中存在信号泄漏和去条带噪声不明显的问题,提出对球谐位系数主成分分析的改进方法。首先对球谐位系数进行主成分分析,提取主要的模态,再对不同模态根据其自身噪声特点选择合适的滤波方法和参数,最后在此基础上进行信号提取。以CSR发布的2005-01~2013-12(108个月)GRACE时变地球重力场模型进行实验,提取2006-06和2006-12等效水高数据和改进前的主成分分析进行比较,表明此方法在有效地剔除条带误差的同时又使得信号泄漏减小。

According to the problem that signal leakage exists and noise reduction by de-striping is not sufficient in the approach of principal component analysis, we propose an improved method of extracting the signal using the principal component analysis directly from spherical harmonic coefficient. First, the principal component analysis is performed on the spherical harmonic coefficients to extract the main modes. Then, the appropriate filtering method and parameters are selected according to their own noise characteristics in different modes. Finally, the signal extraction is carried out. The GRACE temporal earth gravity field models of 2005-01~ 2013-12 （108 months） released by CSR are used to extract the equivalent water height signals, and the data of June and December 2006 are compared with unmodified principal component analysis. The results show that the modified method is effective and could both eliminate striping noise and reduce signal leakage.