Regularized variance component estimation method of MSBAS InSAR deformation inversion

The Multi-Dimensional Small Baseline Subset Interferometric Synthetic Aperture Radar (MSBAS InSAR) technique significantly reduces the visual ambiguity caused by the single line of sight by integrating multi-orbit SAR data for joint solution. However, due to temporal-spatial decorrelation, topographic errors, orbital errors and phase unwrapping, different orbit data usually have different error statistical characteristics, and it is unreasonable to set the equal weight for ascending and descending observations in the inversion process of the existing MSBAS InSAR deformation model. In addition, the close acquisition times of ascending and descending data lead to strong multicollinearity among the column vectors of the design matrix, and the coefficient matrix of the non-regularized MSBAS solution exhibits a severe ill-posedness, with a condition number on the order of 1018, which makes the solution extremely sensitive to noise and results in unstable weight estimation. Hence, MSBAS InSAR deformation modelling method based on regularized variance component estimation (MSBAS-RVCE) is proposed, in which the regularization method is used to ensure stable model solutions, while iterative refinement of the weight matrix provides accurate posterior variances and enhances result reliability. The proposed method was applied to MSBAS InSAR deformation monitoring in the Jiyang mining area. Comparative analysis of subsidence values, profiles, levelling data and InSAR results demonstrates that the proposed method reasonably determines the weight ratio of ascending and descending data in the joint adjustment and improves the accuracy of the results by approximately 35% compared with the equal-weight strategy.