The Influence of Topography-Dependent Atmospheric Delay for the InSAR Time-Series Results and the Deep Neural Network Correction

Topography-dependent atmospheric delay (TDAD) plays a crucial role in limiting the accuracy of deformation monitoring in interferometry synthetic aperture radar (InSAR) measurements. Although several correction methods have been proposed to achieve satisfactory results in differential InSAR (DInSAR), the impact and analysis of TDAD in time-series InSAR have often been overlooked. This study focused on the landslide monitoring near the Baihetan hydropower station located in the rugged and steep terrain of southwestern China. Utilizing C-band Sentinel-1A satellite data, we systematically examine and analyze the influence of TDAD on time-series InSAR. We introduce a deep neural network (DNN) correction model to remove the TDAD, resulting in a substantial improvement in the outcomes of time-series InSAR. The validation proves the effectiveness of this correction, providing valuable insights and technical support for precise TDAD correction in future time-series InSAR applications.