Abstract:[Objective] The topographic characteristics of abandoned farmland in mountainous areas were analysed in order to provide a scientific reference for regional land resource management and sustainable agricultural development. [Methods] Taking Wushan County of Chongqing City as the research area, this study used Landsat TM/ETM+/OLI and Sentinel-2 data on the Google Earth Engine (GEE) cloud platform and adopted decision tree and time series normalised difference vegetation index (NDVI) change detection methods. We extracted and analysed abandoned land information from the study area from 2017 to 2021. [Results] ① From the perspective of time series characteristics, the area of abandoned land in the study area showed an overall upward trend; the increase from 2017 to 2021 was 2 123.50 hm2, with a growth rate of 19.61%, and the curve shape within the interval showed a “W” shaped characteristic. Spatially, abandoned land was regionally dispersed and locally concentrated. It was mainly distributed along the direction of the water system, significantly concentrated on both sides of the river, and surrounded by sloping farmlands. ② The abandoned land in the study area was distributed differently across different elevation and slope zones. The abandoned land was concentrated at elevations below 1 000 m and slopes within a range of 5°~20°. In areas with an elevation below 1 500 m, the area of abandoned land and rate of abandonment showed a pattern of “first increasing and then decreasing,” reaching the highest point in 2019; in areas with an elevation above 1 500 m, the area of abandoned land and the rate of abandonment showed a pattern of “first decreasing and then increasing” over time. The law of “increase” reached its lowest value in 2020. ③ The distribution index of abandoned land under different terrain levels was continuously decreasing, and the distribution index of terrain gradients 1 and 2 was greater than 1, indicating the dominant area of abandoned land. [Conclusion] The combination of a decision tree and NDVI time series change detection methods can accurately identify abandoned land with a recognition accuracy of 83.59%.