Coherence Optimization in Polarimetric Subspace of Sentinel-1 for Subsidence Monitoring

The main purpose of this paper is to investigate the potential of different coherence optimization methods when applied to Sentinel-1 images with the polarimetric subspace of VV and VH channels. These methods try to solve the optimization problem by considering different assumptions including either considering single or multiple scattering mechanisms for an image pair. Moreover, in the case of a large number of SAR images, an estimate for the optimal scattering mechanism can be obtained either separately for every single interferogram or all of the interferograms as a whole, namely, as single- or multi-baseline methods, respectively. On this basis, four different optimization methods including single-baseline multiple scattering mechanism (SB-MSM), single-baseline single scattering mechanism (SB-SSM), multi-baseline single scattering mechanism (MB-SSM), and a revision on the latter one (R-MB-SSM) were applied on 24 interferograms processed over Marvdasht plain (Iran) where is subject to land subsidence. The performance of the optimization methods was evaluated based on the amount of improvement made on the coherence. The results showed that the single-baseline single scattering mechanism method had a high superiority in coherence maximization, but the overall improvement was not considerable. Around 14.91% of the study area showed the coherence improvement of more than 0.1 when using SB-SSM while this percentage was about 14.8, 4.5, and 4.72% for SB-MSM, MB-SSM, and R-MB-SSM, respectively. The low performance of the coherence optimization on Sentinel-1 was caused by insignificant contribution of VH to the optimization due to its high decorrelation property. Moreover, lack of HH polarization in Sentinel-1 images alongside with channel VV prevented the coherence from more enhancement. Further investigation on modification of the phase quality was performed on the time-series analysis results. The greatest number of pixels was subject to small changes in deformation rate after optimization. The maximum amount of root-mean-square difference (RMSD) between the subsidence rate map obtained from the optimization and before that was 0.0086 m/yr which was corresponding to SB-MSM method. This illustrated the most modification to interferometric phase after optimization which was mostly occurred in the areas with lower degree of coherence, i.e., vegetated areas.

Automated summary

This paper investigates the potential of different coherence optimization methods applied to Sentinel-1 images and evaluates their performance. The results show that the single-baseline single scattering mechanism method has advantages in coherence maximization, but the overall improvement is not significant. The low performance of coherence optimization on Sentinel-1 is mainly due to the weak contribution of the VH channel and the lack of HH polarization.