Abstract: Based on the Landsat remote sensing data from Google Earth Engine platform, the mean value algorithm method was used to construct a high spatiotemporal resolution dataset which extracted the information of rocky desertification patches and analyzed the spatio-temporal evolution process of rocky desertification in Qujing. The results show that: This method can effectively reduce the amount of data download and computation while reducing the impact of vegetation seasonal changes on rocky desertification patch information extraction and improving data comparability by using the GEE platform and its image processing. The spatial distribution of rock desertification shows a horizontal distribution pattern of more in the northern than in the southern and serious in the western than in the eastern. Meanwhile, the vertical distribution pattern was that the slope was positively correlated with the area of rocky desertification occurrence and evolution magnitude in areas where the slope was below 15°, and the opposite in areas above 15°. The evolution of rocky desertification from 2000 to 2020 could be divided into five stages: unstable, obvious improvement, serious deterioration, continuous improvement and stable improvement. In the past 21 years, the occurrence of rocky desertification has been reduced by 3390.42 km² (accounting for 11.73% of Qujing's land area), with an average annual reduction of 161.45 km²(accounting for 0.56% of Qujing's land area). The quality of ecological environment in the study area has been improved.