Landcover dataset at Qilian Mountain area (V1.0) (2018)

This dataset contains land cover products in Qilian Mountain Area in 2018. The dataset was produced by two steps. Firstly, land cover product in 2015 is produced using time series Landsat-8/OLI data. In view of the different NDVI time series curves of various land features with time variation, the knowledge of different land features is summarized, the extraction rules of different land features are set, and the land cover classification map in 2015 is obtained. The classification system refers to IGBP and FROM_LC classification system. It is divided into 10 categories: cultivated land, woodland, grassland, shrub, wetland, water body, impermeable surface, bare land, glacier and snow cover. According to the accuracy evaluation of Google Earth high-definition image and field survey data, the overall accuracy of land cover classification products in 2015 is as high as 92.19%. Secondly, taking the land cover classification products in 2015 as the base map, a large number of samples are selected according to the proportion of different types. Based on the Landsat series data and powerful data processing ability of Google Earth Engine platform, the random forest classifier is selected to train the band information and NDVI, MNDWI, NDBI and other indices by using the idea of in-depth learning. The land cover of 2018 is produced. It is concluded that the land cover classified products based on Google Earth Engine platform have good consistency with those based on time series method. In conclusion, the land cover data set in the core area of Qilian Mountains has high overall accuracy, and the method based on sample training of Google Earth Engine platform can expand the existing classification products in time and space, and the frequency of every five years can reflect more land cover type change information in long time series.