A Method by using predefined windows to improve the InSAR atmospheric correction model based on GNSS ZTD and its horizontal delay gradient

Abstract

ABSTRACT Correcting the signal delay caused by electromagnetic waves passing through the atmosphere is crucial in InSAR. Atmospheric delays are significant enough to obscure the true deformation signal, making it difficult for InSAR to detect deformation signals at the millimetre scale. Therefore, it is imperative to mitigate atmospheric interference appropriately. To interpolate sparse GNSS networks and capture small-scale atmospheric delays, this study presents an enhanced method for constructing an atmospheric correction model based on GNSS Zenith Total Delay (ZTD) and its horizontal gradient, which is integrated within a predefined window. To validate the effectiveness of the improved model, we applied it to 27 InSAR short-time and short-spatial baseline unwrapped interferograms. These interferograms are composed of images observed by the Sentinel-1A satellite from January to December 2022 in Southern California. The corrective impact of the enhanced model is also compared against that of the initial inversion model and GACOS to confirm the viability of the approach. Results of the evaluation, considering root mean square error (RMSE), standard deviation (STD), and the correlation between phase and elevation, indicate that our enhanced method reduces correction errors by 35.10%, 20.05%, and 63.52%, respectively, compared to GACOS. Compared to the initial model, these reductions are 58.19%, 57.23%, and 4.34%, respectively. By integrating the outcomes from all three evaluation methods and analysing the corrected images, our enhanced model effectively resolves the inconsistency in correction effects resulting from the spatial variability of atmospheric delay. Moreover, incorporating the horizontal gradient in the inversion method has proven effective in capturing small-scale atmospheric delay signals. This highlights the feasibility and benefits of our enhanced model.