GRACE时变重力位系数误差处理滤波方法研究

Study on filtering methods of error processing in GRACE time variable gravity data

由于GRACE时变重力位系数中高阶次项存在较大的误差,奇(偶)项阶数之间存在着相关性误差,直接利用GRACE时变重力场模型数据反演地表质量变化时,会使结果出现严重的条带噪声,必须采用最佳滤波函数进行空间平滑.本文从滤波理论出发,依据信噪比最大准则和模拟误差比较分析了不同滤波方法的优缺点,并以华北平原为例进行了数值分析.实验表明,各向异性滤波处理结果优于各项同性滤波;去相关滤波中PnMl方法处理结果优于滑动窗口去相关;单一滤波在处理结果上不是很理想,采用组合滤波可以提高反演信噪比和空间分辨率;对于滤波处理带来的信号泄露等误差可以运用尺度因子法进行有效改正.

The spherical harmonics of time-variable gravity models have correlation problem and measurement errors in high orders and degrees, the water storage variation inversion results would be filled with serious north-to-south stripe when using time-variable gravity data directly. Thus, filtering is needed to remove them before using the Gravity Recovery and Climate Experiment(GRACE) time-variable gravity models. Based on the filtering theory, this paper analyzes the advantages and disadvantages of different filtering methods in terms of signal-to-noise ratio criterion and simulation errors, and makes a numerical analysis with the North China Plain as an example. The experimental results show that the processing results of non-isotropic filtering are better than isotropic filtering. The processing results ofPnMlmethod in decorrelation filtering are better than sliding window decorrelation. The result of using only one filtering method is not very ideal. Combination filtering can improve signal-to-noise ration and spatial resolution of the inversion. The scale factor method can be used to effectively correct errors such as signal leakage caused by filtering processing.