Identification of waterlogging in Eastern China induced by mining subsidence: A case study of Google Earth Engine time-series analysis applied to the Huainan coal field

Abstract

Underground coal mining with high groundwater levels causes many environmental problems, one of the main ones being subsidence waterlogging. After the subsidence waterlogging, the land faces many problems, such as insufficient land development and difficult and costly land restoration. Therefore, it is necessary to monitor the spatial range and temporal trajectory pattern of surface subsidence in a timely manner. When information about underground mining is lacking, it is difficult by using remote sensing alone to identify and distinguish (i) natural water, (ii) engineering water, and (iii) subsidence waterlogging. In this study, we used the Google Earth Engine platform to develop a method to detect subsidence waterlogging area and the disturbance year. The method includes pixel-based trajectory extraction and object-based water type recognition. First, LandTrendr is used to extract the change water (engineering water and subsidence waterlogging) pixels and its disturbance year. Then, the morphological method to further eliminate engineering water patches, so as to extract subsidence waterlogging. We selected the Panxie mining area in Huainan, China as the study area. Using 33 years of Landsat time-series data to generate values of the annual water frequency index, maps of the year of water accumulation caused by underground coal mining and the year of restoration during 1989–2016 are drawn with accuracies of 86.5% and 80.7%, respectively. The results show that from 1989 to 2016, the accumulated area of subsidence waterlogging was 7715.25 ha, accounting for 14.5% of the total area of the study area, of which 75.8% occurred from 2008 to 2016. Furthermore, the accumulated area of restoration was 207.18 ha, which occurred after 2007 and accounts for 2.7% of the total area of subsidence waterlogging. Based on the analysis results of the changes of waterlogging types, the best time window for restoration of the waterlogging area is 3y after water accumulation. The main innovation of this paper is to make use of the temporal heterogeneity (Number of year in patches, Y_num) and morphological index (Landscape shape index in patches, LSI) of the change water to distinguish engineering water and subsidence waterlogging.