 The study proposes a new method to automatically characterize vegetation degradation and recovery trajectory for multiple mine sites in Beijing, China using Landsat satellite observations on the Google Earth Engine platform. The method constructs temporal compass sites of the normalized differential vegetation index, NDVI, identifies the year of change and pre- and post-change trends using Pettit test and CEN plus Mankendl analysis, classifies time series NDVI into five vegetation trajectory types and 13 subtypes, and analyzes the recovery status of mine sites. The method achieved satisfactory accuracies in trajectory type classification with an overall accuracy of 91.1%. The results showed that our method had better capability in identifying gradual recovery or degradation compared to the land-trender algorithm and yielded much less omission and commission errors of R, D, RD, and DR types. In the study area, 1,469.07 hectares out of 3,746.25 hectares of the mine sites have been recovered from 2000 to 2019, with recovery mainly occurring during 2009 to 2013 when the Green Mine Construction Campaign was launched in Beijing. The study's results are expected to support evaluation of mine restoration effects and detection of illegal mine sites. This article was authored by Wuhan, Yinghaiku, Lijuanxiu, and others. We are article.tv, links in the description below.