A thorough accuracy assessment of MT-InSAR for soil erosion monitoring in the hilly and gully Loess Plateau

Abstract

A reliable monitoring of erosion over large areas is crucial for the control of soil erosion. Spaceborne remote sensing techniques (e.g. Interferometric synthetic aerture radar, InSAR), provide a promising tool for the large-scale erosion monitoring, while their accuracy was rarely quantitatively assessed. In the study, the accuracy of InSAR for erosion mointoring was quantitatively assessed in a small catchment of the hilly and gully Loess Plateau. Twenty-one Sentinel-1 SAR images during July 2021-June 2022 were acquired and analyzed using the Multi-Temporal InSAR (MT-InSAR) technique, while topographic features derived from high-resolution DEMs, generated based on point clouds acquired by umanned aeria vehicle laser scanning (ULS), was employed to achieve land surface deformation related to erosion and deposition. The erosion / deposition results obtained were then verified with erosion pin measurements, differencing results of ULS-derived DEMs (DoD), and precipitation records. Results showed that average erosion rate of the study area detect by InSAR (9882 t km⁻² a^-1) was similar to that derived by ULS DoD (9747 t km⁻² a^-1). The spatial pattern of erosion / deposition derived by MT-InSAR exhibited a generally good agreement with that of DoD results, while, unlike ULS-derived DoDs, InSAR failed to detect changes of narrow gullies. The derived erosion and deposition were significantly correlated with erosion pin measurements (R² = 0.85, p < 0.01, ρ = 0.92), with an RMSE of 6.51 mm, when two sites with decorrelation phenomena were not considered. The geomorphic change of gully heads / gully slopes detected by MT-InSAR showed a significantly positive correlation with precipitation (ρ > 0.85, p < 0.05). Overall, we concluded that MT-InSAR can be used for long-term erosion monitoring over a large area, while an integration of multi-source data (e.g. optical images) was still desirable to overcome the drawbacks of InSAR technology such as those induced by the decorrelation phenomena on areas with large and / or rapid changes.