Fine‐scale accuracy assessment of the 2016 National Land Cover Dataset for stream‐based wildlife habitat

Abstract

AbstractEfficiently and effectively identifying and assessing potential wildlife habitat and important ecological resources is essential as rapid anthropogenic land use change alters and detrimentally affects terrestrial and aquatic habitats. Accuracy assessment of remotely sensed data supports ecological planning and management decisions, and is especially important when using freely available, coarse‐resolution spatial datasets, such as the National Land Cover Dataset (NLCD). A popular dataset designed for application at larger regional to national spatial scales, the NLCD has been used in finer scale studies outside of its intended use, often without the imperative support of field‐based accuracy assessment. We ground‐truthed stratified random sampling points to assess the accuracy of the 2016 NLCD at a fine spatial scale relevant to stream‐based ecological research. Our results demonstrated an overall accuracy of <65%, less than the United States Geological Survey recommended accuracy of ≥85%. Results indicated that the NLCD may not be effective as a tool for stream‐level studies and could provide erroneous results for fine‐scale habitat assessment and planning when used as the only reference dataset. When conducting ecological research, it is important to consider the appropriate scale, resolution, and limitations of available datasets to achieve the most accurate results.