This dataset represents the cumulative result of multi-season classification of land cover in the GCPO LCC geography to NatureServe Ecological Systems based on 2011 seasonal Landsat Satellite Imagery. The approach used a Random Forest algorithm and several dozen input data layers to classify land cover at a 30 m pixel resolution. The description below is taken directly from the report titled “Update of the Eastern GCPO Land Cover Database to 2011 Using a LS2SRC Approach”, by Dr. Qingmin Meng, Department of Geosciences, Mississippi State University.Random Forest classifier is based on the general decision tree approach, which has been a popular approach to multilevel and multistage decision making. Its basic idea involves breaking up a complex decision into a union of sequenced simplified decisions with the assumption that this method will result in the expected outcome. No field data is used in this classification. The training data are randomly generated based on the updated national GAP data for the east GCPO region. 500 sampling sites of each type of ecological system were randomly created from its large habitats. Then, two graduate students using Google Earth manually check from site to site to determine whether a sampling site is the type of ecological system as indicated in GAP dataset. After a visual assessment, the sample size of each ecological system is about 350, which was used as input training data for classification. After classification, 100 sampling sites of each ecological system were then randomly and independently generated for classification evaluation. Likewise, a similar visual checking process using Google Earth was performed by the two graduate students to remove any unreliable locations. Then, the rest samples are used to assess classification accuracy. The data products include the intermediate data products and the final data product ecological systems of east GCPO region and includes a total 62 geospatial layers covering the east GCPO region. The seamless raster layers of Landsat spring, fall, and winter images, the derived remote sensing indices and TC transformation, landform indices, and the seamless soil series map covering the whole east GCPO data can be available for free at the GCPO data server for any interested parties; these data are also available for free by contacting Meng (email@example.com) at Mississippi State University. The 2011 ecological systems data and these geospatial data provided the local governments, communities, nonprofit organizations, and researchers with the updated and enhanced geospatial information for natural resources management, landscape planning, and ecological conservation. We used the methods of NLCD Mapping Tools, SVM, NNC, and RFC. However, the results from NCLD Mapping Tool, SVM and NNC were much poorer than RFC. The 117 classes and their ecological system names are summarized in Table 3, and the accuracy is summarized in the Table 4 of the project final report. The average user’s accuracy is 60.33%, the average producer’s accuracy is 60.35%, and the over accuracy is 60.33%. This classification is comparable to other regional classifications, for example, the Southwest Regional Gap Project, in which 85 classes evaluated with overall accuracy 61% and 66% for four forest cover attributes in Cohen and Goward’s (2004) study in western Oregon. Comparing the user’s and producer’s accuracy with the GAP Analysis conducted by Lowry et al. (2007), this ecological systems analysis produced relatively better results.