Continues monitoring of subsidence water in mining area from the eastern plain in China from 1986 to 2018 using Landsat imagery and Google Earth Engine

The eastern plain of China is one of the most important grain production areas in China. Meanwhile, the plain is also an important coal production area with a large number of “coal-grain composite regions”. Coal mining causes land subsidence and waterlogging, which destroys a significant amount of cultivated land. However, there is no dataset for the spatio-temporal distribution of subsidence water to assist related decision-making on a regional scale. Accurate monitoring of subsidence water is still a challenge, especially distinguishing among subsidence water, natural water and artificially excavated water by only using remote sensing data. Here, A new method to generate a dynamic map for the subsidence ponding year and the restoration year using the Google Earth Engine platform with 33-year-old Landsat imagery was created. The time segmentation method was used to first extract the change water pixels corresponding to subsidence water and artificially excavated water with similar trajectory features. Then, the morphological characteristics of the two types of change water at the beginning year of water accumulation are used to construct 13 landscape indexes. This approach utilizes the Random Forest algorithm to distinguish between subsidence water and artificially excavated water. The Huang-Huai-Hai plain area in eastern China was selected as the study area and extracted the subsidence water areas from 1986 to 2018. The identification accuracies for subsidence ponding year and restoration year are 88% and 85%. 79% of the subsidence water areas are located in cultivated land, which shows significant impacts to agricultural activities. The method proposed could be applied to other similar areas, the results could provide reference and data for decision-making and related land reclamation planning. •A novel methodology was developed for mapping mining subsidence water.•Map of the subsidence ponding year and restoration year.•Random forest algorithm can effectively extract the patch of subsidence water.•Contribute to monitor annual subsidence water at a regional scale.•This paper belongs to change detection using Landsat time series.