Szczegóły publikacji
Opis bibliograficzny
Hydro-spectral Sentinel-1/2 precursors of sinkhole occurrence in a flooded post-mining area / Wojciech T. WITKOWSKI, Artur GUZY, Magdalena ŁUCKA, Krzysztof Kusztykiewicz // IEEE Geoscience and Remote Sensing Letters ; ISSN 1545-598X . — 2026 — vol. 23 art. no. 8001005, s. [1–5]. — Bibliogr. s. [5], Abstr. — Publikacja dostępna online od: 2025-12-17
Autorzy (4)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 165706 |
|---|---|
| Data dodania do BaDAP | 2026-03-16 |
| Tekst źródłowy | URL |
| DOI | 10.1109/LGRS.2025.3645601 |
| Rok publikacji | 2026 |
| Typ publikacji | artykuł w czasopiśmie |
| Otwarty dostęp |  |
| Creative Commons | |
| Czasopismo/seria | IEEE Geoscience and Remote Sensing Letters |
Abstract
Sinkholes generated above flooded underground mines pose a growing hazard, yet practical precursors remain lacking. We analyze five years (2018–2023) of Sentinel-1 C-band radar and Sentinel-2 multispectral imagery over the abandoned Olkusz–Pomorzany underground zinc–lead mine (Poland), where 19 sinkholes appeared during groundwater rebound in early 2022. Three hydro-spectral proxies (HSPs), C-band backscatter coefficient, moisture index (MI), and normalized difference vegetation index (NDVI), were extracted for sinkhole pixels and 500 randomly distributed background control pixels. A control–minus–event difference isolates local effects from regional variability. Welch t, Mann–Whitney U, and bootstrap tests all indicate highly significant post-collapse mean shifts in the three HSPs. Breakpoint analysis applied to the 26-month pre-event record reveals a common structural change in mid-2021, approximately six months before the first sinkhole and coincident with rapidly rising groundwater. The Chow test confirms a significant difference in regression coefficients across the detected break. These results demonstrate that freely available hydro-spectral data can supplement synthetic aperture radar interferometry-based deformation measurements, offering weeks to months of lead time for early warning in vegetated, post-mining terrain. The approach is inexpensive, transferable, and readily automated within Google Earth Engine (GEE) and R.
