How Can SBAS‑InSAR and KTree‑AIDW Boost Mining Subsidence Monitoring?

Coal mines often cause ground subsidence, a hazardous phenomenon for surface structures. To keep these movements in check, scientists use SBAS‑InSAR, a radar technology that measures deformations across large areas with centimeter precision. The catch? When the radar signal gets "confused" by vegetation or surface changes, crucial data are lost.\n\nThe new study merged SBAS‑InSAR with KTree‑AIDW, an interpolation algorithm that searches for the most similar points around a missing area and assigns them different weights based on their homogeneity. In practice, it’s like filling puzzle gaps with the best‑fitting pieces.\n\nUsing Sentinel‑1 images with a 12‑day revisit cycle, researchers examined the grouting front of the Liangbei mine in China during 2021. Results show a 0.95 correlation with traditional measurements, an average error of just 22 mm, and markedly better accuracy in decoherence zones. The cumulative maximum subsidence reached 233 mm, with an average rate of 171 mm per year, yet all values stay within Class I damage thresholds.\n\nThis technique combo paves the way for more reliable monitoring, even under tough conditions, and could become the global standard for mining safety.

SBAS‑InSAR (Small Baseline Subset Interferometric Synthetic Aperture Radar) exploits a series of radar images acquired by satellites such as Sentinel‑1 to build interferograms with a small temporal baseline. This choice reduces temporal decorrelation effects but does not completely eliminate decoherence zones, typically generated by vegetation changes, soil moisture, or back‑filling activities.\n\nKTree‑AIDW (K‑Tree Adaptive Inverse Distance Weighting) is a spatial interpolation algorithm that organizes sampled points into a balanced tree structure (K‑Tree). This structure enables rapid neighbor searches while accounting for local data density variability. The “adaptive” aspect stems from assigning weights inversely proportional to distance, modulated by an heterogeneity factor that penalizes samples from regions with markedly different statistical characteristics. In practice, the method reduces outlier influence and improves continuity of the interpolated deformation field.\n\nIn the Liangbei mine case study, Sentinel‑1 images (C‑band, 5 × 5 km² coverage) were pre‑processed with the ESA SNAP pipeline for co‑registration, topographic removal (DEM), and interferogram generation. SBAS produced a temporal series of vertical displacement maps, but with gaps corresponding to strong decoherence areas. KTree‑AIDW then filled these gaps, generating a continuous subsidence surface.\n\nValidation metrics include the correlation coefficient (R = 0.95), RMSE (22.08 mm) and RRMSE (9.48 %). Analysis of characteristic points in decoherence zones showed an MAE of 19.05 mm, indicating improvement over traditional interpolation methods (e.g., simple IDW or kriging). Moreover, curvature and slope modeling allowed assessment of compliance with Class I structural damage thresholds, with maximum inclinations of ±2.9 mm/m and curvatures of ±0.18 mm/m².\n\nFrom a computational standpoint, the SBAS‑InSAR + KTree‑AIDW integration requires an intensive preprocessing phase (co‑registration, speckle filtering) followed by a local interpolation that, thanks to the K‑Tree structure, scales almost linearly with the number of sampled points. This makes the method suitable for large‑scale, high‑frequency monitoring, paving the way for early‑warning systems in underground mining and other geotechnical engineering contexts.