Improved Landslide Monitoring in Low-Coherence Mountainous Areas: A Coherence-Enhanced Multi-Temporal InSAR Approach

High-precision measurement in low-coherence areas remains posing challenges for multi-temporal synthetic aperture radar interferometry (InSAR). For instance, over the regions covered by dense vegetation, InSAR merely provides sparse measurement points (MPs) due to high spatio-temporal decorrelation. To address this, our study proposes a coherence-enhanced multi-temporal InSAR (CE-InSAR) approach, to better monitor low-coherence landslide displacement in the radar line-of-sight (LOS) direction. The key ideas of CE-InSAR include the pre-processing feasibility assessment for obtaining the important pre-defined parameters for time series processing and coherence enhancement with the use of phase optimization for the C-band Sentinel-1 data stacks. To demonstrate the effectiveness of CE-InSAR, 85 scenes of Sentinel-1 images (2021-2023) covering the Tianxi landslide in Guangxi Province, China with an NDVI value greater than 0.5, were applied to retrieve the historical displacements and analyze the activity state. The InSAR measurements both from the pre-sling and post-sliding phases illustrated the significant advantages of CE-InSAR, with five more times of MPs both in a single landslide scale and regional scale, compared to small baseline subset interferometry (SBAS-InSAR). Furthermore, time series analysis considering rainfall factors, indicates CE-InSAR can detect the accelerated displacement of the Tianxi landslide prior to sliding, exhibiting a maximum LOS accumulative displacement of around 70 mm. Subsequently, the combining impacts of human activity and rainfall contributed to the landslide's failure. Finally, the major uncertainties and limitations for the application of CE-InSAR were discussed and the conclusions were summarized. In general, this research is valuable and useful for guiding landslide displacement monitoring characterized by low coherence using InSAR.