A robust adaptive InSAR-GNSS fusion approach for three-dimensional surface deformation monitoring

Abstract

Urban land subsidence threatens coastal infrastructure, especially in reclaimed and densely developed areas. Single-geometry InSAR (Interferometric Synthetic Aperture Radar) provides wide-area LOS (line-of-sight) deformation but is limited in resolving 3D (three-dimensional) ground motion. This study develops a robust InSAR-GNSS (Global Navigation Satellite System) fusion framework for three-dimensional deformation monitoring in Hong Kong. Sentinel-1A images from January 2020 to April 2024 were processed using the ISCE2-MintPy SBAS-InSAR (Small Baseline Subset-InSAR) workflow, while GNSS observations were resampled and interpolated to provide east, north, and vertical constraints. An Iteratively Reweighted Least Squares-Helmert Variance Component Estimation method, termed IRLS-HVCE, was proposed to combine robust outlier suppression with adaptive variance-component weighting. Independent GNSS validation shows that IRLS-HVCE achieves the lowest combined three-dimensional RMSE (Root Mean Square Error) of 9.87 mm/yr. The framework provides a practical approach for three-dimensional deformation monitoring in coastal cities.