A Distributed Scatterer InSAR Framework for Long-Term Monitoring of Soil Erosion Dynamics over Agricultural Landscapes Using Sentinel-1

Soil erosion is a major driver of land degradation in cultivated landscapes, but regional-scale monitoring of long-term surface change remains difficult. Interferometric Synthetic Aperture Radar (InSAR) can detect millimetre-scale deformation, yet conventional Persistent Scatterer and Small Baseline Subset approaches are poorly suited to cropland because tillage, crop development, residue management, and soil-moisture variability cause severe temporal decorrelation.

This study develops a Distributed Scatterer InSAR framework for long-term monitoring of agricultural surfaces using Sentinel-1. The method was tested in the Middle Volga region of European Russia using 222 Sentinel-1A IW SLC VV scenes acquired between May 2016 and October 2025. A conventional ISCE2-MintPy SBAS workflow produced a median temporal coherence of only 0.03 over cropland, and fewer than 0.4% of pixels met the usual coherence threshold, demonstrating the practical failure of Persistent Scatterer-based processing in intensively cultivated fields.

The proposed framework replaces stable-point referencing with independent pixel-wise temporal integration of short-baseline interferograms. It combines one-dimensional temporal phase unwrapping, phase-closure quality control, ERA5-based tropospheric correction, SMAP-based soil-moisture compensation, and a seasonally optimized interferogram network designed to reduce vegetation-related decorrelation. The final network included 168 selected interferometric pairs with perpendicular baselines below 150 m.

The method yielded valid deformation estimates for approximately 809,000 ha of cropland, corresponding to 53.5% of the agricultural mask, at an effective spatial resolution of about 60 m. The median long-term vertical displacement proxy was +2.70 mm yr⁻¹, while 44.6% of the valid area showed negative values. However, the median seasonal amplitude reached 23.2 mm, approximately fifteen times greater than the long-term trend. This indicates that agricultural operations, crop rotation, and seasonal soil-moisture changes dominate the observed radar signal.

These results show that agricultural dynamics must be treated as an explicit component of the InSAR observation model rather than as residual noise. Integration with annual crop maps, rainfall erosivity, and field laser-scanning data is therefore required to isolate the geomorphic component and enable robust regional assessment of erosion and accumulation.