|Land cover mapping via remote sensing is an important tool for conservation and land management. A critical component to land cover mapping is defining the classification. A classification scheme must be sufficiently detailed to meet the goals to which the map will be applied yet simple enough to accurately map the classification units with the available data and classification methods.
This thesis describes the methods and presents the results and accuracy assessment of a map of NatureServe’s Ecological Systems in the East Gulf Coastal Plain, USA derived using Landsat ETM+ imagery. A combination of remote sensing techniques and classification methods was used to generate a 50 class land cover map.
Of 43 Ecological Systems existing in the East Gulf Coastal Plain, 25 were mapped and an additional 8 modified system classes were mapped. The remaining 17 mapped land cover classes were composed of anthropogenic classes and land cover lacking vegetation.
Additionally, an accuracy assessment of the land cover map was performed and interpreted by assessing the causes of error. In this land cover map, a majority of the errors are caused by either fuzzy boundaries between class definitions or a lack of spatial data that can reliably separate classes. This error analysis provides insight into the utility of the classification scheme when mapping with remotely sensed data.