Correction of spatially varying stratified atmospheric delays in multitemporal InSAR

Abstract

Tropospheric delay is a major limiting factor in ground displacement retrieval using interferometric synthetic aperture radar (InSAR). Such delays are caused by the variations of pressure, temperature, and humidity of the troposphere, potentially leading to elevation-dependent phases in interferograms. The atmospheric heterogeneity makes the phase-elevation dependence vary in both space and time, posing challenges on empirical models that estimated the elevation-topography relationship in empirically predefined windows interferogram by interferogram. Moreover, the portion of the deformation that is correlated with topography can be mistaken as tropospheric delays and therefore degrade the performance of linear regression between the tropospheric phase and the elevation. The effectiveness of these methods is further limited by the requisite knowledge presumed of the user, particularly with regard to how to divide the scene and which observation model is suitable. These limitations make reliable isolation of tropospheric delay challenging. We present here a novel approach that combines tropospheric delay, deformation and possible topographic error to a joint model, enabling a precise separation of tropospheric delay. The rationale is rooted in the distinct temporal dependencies of these parameters. To avoid empirical setting of segmentation windows, we employ quadtree to adaptively control the spatial variability of tropospheric properties. We validate the proposed method at two volcanic areas, Bali and Hawaii, where tropospheric delay presents notable variability coupled with elevation-correlated deformation signals. Both ascending and descending Sentinel-1 data over the island of Bali show that the proposed method outperforms the conventional ones. The misfit standard deviation (STD) is reduced by ∼50% at the island of Bali, where elevation-correlated deformation is simulated according to a predefined model. More impressively, at the island of Hawaii, the misfit STD between InSAR and ground truth (i.e., GPS displacements) decreases from 25.1 to 4.5 mm. The experimental results demonstrate that the proposed method is effective in reducing the spatially variable tropospheric delay while preserving the elevation-correlated deformation signal.