Landsat-derived annual maps of agricultural greenhouse in Shandong province, China from 1989 to 2018

Using 8,450 Landsat images on the Google Earth Engine, we built the first Landsat-derived annual maps of agricultural greenhouse (AG) in Shandong province, China from 1989 to 2018. Two types of reference datasets, including AG and Non-AG, were labeled via visual inspection of high-resolution imagery available in Google Earth or Landsat imagery based on a 10 km grid sampling structure. The mapping window for each year was selected based on the vegetation growth and the phenological information. Classification for each year was carried out initially based on the random forest classifier after the feature optimization. A temporal consistency correction algorithm based on classification probability was then proposed to the classified AG maps for further improvement.