Sinkhole occurrence monitoring over shallow abandoned coal mines with satellite-based persistent scatterer interferometry

Unexpectedly occurring sinkholes caused by shallow submerged voids are one of the key problems of intensively built-up post-mining areas. Five percent of Poland's territory is still under the influence of post-mining deformation. Impending sinkholes, which tend to develop randomly over very large areas, cannot be detected by using traditional geophysical methods. Moreover, geodetic measurement methods like Global Positioning System (GPS) analyses, tachymetry, laser scanning, or photogrammetry may be useful only for the registration of the dimensions and locations of sinkholes that have already occurred. Here, we investigate an area in Upper Silesia, Poland, where 345 sinkholes were recorded over a period of more than 20 years (1992-2013). Most of the events occurred in intensively built-up areas, and thus, these sinkholes posed a direct threat to the population. In the test area, 11 sinkholes were detected with a maximum depth of 16 m and maximum dimension of 25 m. The root cause of sinkhole formation was the collapse of post-mining shallow voids. We used satellite radar technology to detect and monitor ground movements potentially associated with the impending sinkholes. The findings showed that the application of Persistent Scatterer Interferometry can support the identification of zones where sinkholes will occur, given adequate spatio-temporal sampling. A stack of Envisat Synthetic Aperture Radar (SAR) images acquired between March 2003 and August 2010 was used in the analysis, and the results confirmed that precursory ground movements were detectable at an early stage of sinkhole development. The land subsidence rate observed prior to sinkhole collapse was not constant in time. Accelerated ground movements within 100 m of an observed sinkhole were detected. We conclude that satellite measurements may provide significant support in the early identification of areas prone to sinkhole occurrence.