发表刊物：Measurement Science and Technology
摘要：The paper proposes a spatiotemporal joint state-space (STSS) model, where the state vector contains only estimates of the true geophysical signals, while the striping errors and high-frequency noise are treated as observation noise, the covariance matrix reflecting the statistical information of stripe errors is used as the observation noise covariance matrix in the state-space model. The state equation is constructed by considering the correlation of the time-varying gravity field at adjacent time nodes, and a covariance matrix of process noise is generated using a power-law model. Based on this, spatial domain constraints are applied through pseudo-observation equations, and variance component factors are introduced to adaptively adjust the size of the noise covariance matrix. The parameter values are determined by minimizing the cost function and optimizing the signal-to-noise ratio. Using the gravity recovery and climate experiment Level-2 spherical harmonic (SH) product as the validation dataset, first, the model is evaluated in terms of the correlation between SH coefficients and mascon, where the STSS model shows a higher correlation compared to other filtered solutions. Secondly, the model’s performance is analysed based on signal and noise levels, demonstrating superior performance compared to the decorrelation and denoising kernel (DDK) filter series and combined filters. Lastly, from the perspective of global and regional surface mass migration estimates, the STSS model shows improvements over the temporal state-space model, with root mean square errors comparable to the optimal DDK filter across various regions. When using the hydrological model as a reference, the STSS model achieves the highest correlation coefficient.
论文类型：期刊论文
是否译文：否
收录刊物：SCI