Land-Cover Controls on the Accuracy of PS-InSAR-Derived Concrete Track Settlement Measurements

Abstract

What are the main findings?

PS-InSAR accurately captured millimeter-scale settlements along the Honam High-Speed Railway embankments, showing strong agreement with leveling survey results (MAE = 1.7–4.2 mm).

Quantitative regression analysis demonstrated that land-cover composition—particularly the balance between vegetation and high-reflectivity surfaces—explains a significant portion of the variability in PS-InSAR accuracy and persistent scatterer density.

What is the implication of the main finding?

The study transforms the well-known limitation of vegetation-induced decorrelation into a predictive framework by statistically modeling its influence on PS-InSAR performance.

The proposed regression-based approach provides practical guidance for selecting monitoring zones and determining when complementary ground-based surveys are required, thereby improving the reliability of satellite-based settlement monitoring strategies for railway infrastructure management.

Accurate monitoring of settlement in high-speed railway embankments is critical for operational safety and long-term serviceability. This study investigates the applicability of Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) for quantifying millimeter-scale deformations and emphasizes how surrounding environmental factors influence measurement accuracy. Using 29 TerraSAR-X images acquired between 2016 and 2018, PS-InSAR-derived settlements were compared with precise leveling survey data across twelve representative embankment sections of the Honam High-Speed Railway in South Korea. Temporal and spatial discrepancies between the two datasets were harmonized through preprocessing, allowing robust accuracy assessment using mean absolute error (MAE) and standard deviation (SD). Results demonstrate that PS-InSAR reliably captures settlement trends, with MAE ranging from 1.7 to 4.2 mm across different scenes. However, significant variability in accuracy was observed depending on local land-cover composition. Correlation analysis revealed that vegetation-dominated areas, such as agricultural and forest land, reduce persistent scatterer density and increase measurement variability, whereas high-reflectivity surfaces, including transportation facilities and buildings, enhance measurement stability and precision. These findings confirm that environmental conditions are decisive factors in determining the performance of PS-InSAR. The study highlights the necessity of integrating site-specific land-cover information when designing and interpreting satellite-based monitoring strategies for railway infrastructure management.