AUTHOR=Pan Jianping , Zhao Ruiqi , Xu Zhengxuan , Cai Zhuoyan , Yuan Yuxin TITLE=Quantitative estimation of sentinel-1A interferometric decorrelation using vegetation index JOURNAL=Frontiers in Earth Science VOLUME=10 YEAR=2022 URL=https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2022.1016491 DOI=10.3389/feart.2022.1016491 ISSN=2296-6463 ABSTRACT=

Sentinel-1A data are widely used in interferometric synthetic aperture radar (InSAR) studies due to the free and open access policy. However, the short wavelength (C-band) of Sentinal-1A data leads to decorrelation in numerous applications, especially in vegetated areas. Phase blurring and reduced monitoring accuracy can occur owing to changes in the physical and chemical characteristics of vegetation during the satellite revisit period, which essentially makes poor use of SAR data and increases the time and economic costs for researchers. Interferometric coherence is a commonly used index to measure the interference quality of two single-look complex (SLC) images, and its value can be used to characterize the decorrelation degree. The normalized difference vegetation index (NDVI) is obtained from optical images, and its value can be used to characterize the surface vegetation coverage. In order to solve the problem that Sentinel-1A decorrelation in the vegetated area is difficult to estimate prior to single-look complex interference, this paper selects a vegetated area in Sichuan Province, China as the study area and establishes two two-order linear quantitative models between Landsat8-derived normalized difference vegetation index and Sentinel-1A interferometric coherence in co- and cross-polarization: When NDVI at extremely high and low levels, coherence is close to zero, while NDVI and coherence show two different linear relationships in co- and cross-polarization in terms of NDVI at the middle level. The models global error basically obeys the normal distribution with the mean value of −0.037 and −0.045, and the standard deviation of 0.205 and 0.201 at the VV and VH channels. The two models are then validated in two validation areas, and the results confirm the reliability of the models and reveal the relationships between Sentinel-1A InSAR decorrelation and vegetation coverage in co- and cross-polarization, thus demonstrating that the NDVI can be applied to quantitatively estimate the InSAR decorrelation in vegetated area of Sentinel-1A data in both polarization modes prior to SLC interference.