Long-term Landsat monitoring of mining subsidence based on spatiotemporal variations in soil moisture: A case study of Shanxi Province, China

Abstract

Ground subsidence caused by mining activities has extensively and chronically affected people living near mining areas. Remote sensing images provide temporally and spatially continuous observations of surface natural resources. In order to accurately understand and continuously observe subsidence areas, this study investigated the spatiotemporal dynamics of mining subsidence using long-term Landsat time-series. The study area was a typical mining region in Shanxi Province, China, with a long history of mining activity. Landsat images from 1986 to 2019, as well as reference data, were used to construct inter-annual time-series trajectories of soil moisture, which were derived from spatial normalizations of the soil moisture monitoring index(SSMMI). The LandTrendr algorithm on the Google Earth Engine platform was then used to extract the subsidence areas and their spatiotemporal dynamics. The results indicate that (i)the map of the spatial distributions of mining subsidence had an overall accuracy of 83.8% and indicate that areas that underwent subsidence were mainly concentrated in mining cites(including Taiyuan, Lvliang, and Linfen); (ii)subsidence before 1990 accounted for 39.6% of the total occurrences and a large proportion(45.18%) of the subsidence was short-term; (iii)subsidence in mountain areas that have long histories of mining and long subsidence durations require further attention by researchers. This study allows us to advance the long-term monitoring of mining subsidence times using variations in soil moisture, combined with moderate resolution satellite data such as Landsat and time-series analysis algorithms.

Keywords

• Mining subsidence
• Soil moisture
• Spatial normalization
• Spatiotemporal variation
• LandTrendr algorithm
• Google Earth Engine