Monitoring and modelling of progressive land subsidence using multi-temporal space-borne remote sensing measurements

In Chiba Prefecture, Japan, land subsidence in sub-urban regions has become an environmental and social issue which insists frequent monitoring for impact assessment. The inception of land subsidence in this region is attributed to natural factors, such as seismic activity and soil layer consolidation, as well as man-made factors, such as ground water and natural gas extraction. However, accurate estimation of the widely and distinctly distributed multiscale subsidence areas becomes very challenging, time-consuming, and labour-intensive by conventional level survey measurements. In this study, we applied and validated a ground subsidence monitoring method for Chiba Prefecture, Japan, using L-band ALOS-2 space-borne Synthetic Aperture Radar (SAR) satellite data. We used Single Look Complex SAR data in StripMap mode with a 3 m resolution, a swath width of 70 km, and repeat-pass acquisition geometry. Continuously acquired ALOS-2 SAR data in both ascending and descending orbital directions, totally 78 scenes were used from 2016 to 2023. The small baseline subset method was used to stack the interferograms and reduce the phase distortions, and convert them into the corresponding vertical displacement for subsidence measurements. The estimated time-series subsidence results were further assessed for areas with high coherency (> 0.6). The estimated annual subsidence rate from SAR-based measurements confirms the existence of certain land areas where the annual displacement exceeds -15 mm per year and their spatial extent. We used the least-squares method for spatial data with four adjustment parameters to improve the overall accuracy of SAR-based subsidence measurements by integrating them with the sparsely distributed 309 level survey locations. The results were categorized into six classes based on the annual subsidence rates, and the root-mean-square error (RMSE) was compared before and after the improvement of subsidence measurements using the proposed method. The results indicate that the RMSE for each subsidence class is below 5 mm, confirming consistency with national accuracy standards for subsidence monitoring.