“Karst” landscapes are characterized by distinctive hydrology and surface features, such as closed depressions, sinkholes, and sinking streams, that result from high rock solubility and well-developed fracture porosity (Ford and Williams, 2007; Weary and Doctor, 2014). Soluble carbonate rocks underlie most of Middle Tennessee, large areas of East Tennessee, and parts of surrounding states. The U.S. Geological Survey conducted a study, in cooperation with the Tennessee Department of Transportation, to create a geospatial dataset of depressions, sinking streams, and their associated watersheds within karst areas of Tennessee and parts of Kentucky, Virginia, North Carolina, Georgia, Alabama, and Mississippi. The purposes of the study were to determine the methods and assessment techniques used to classify depressions in digital elevation models (DEMs) as likely or unlikely to exist; to delineate depressions, sinking streams, and their associated watersheds from multiple elevation data sources and digital hydrographic data within the karst areas of Tennessee and parts of surrounding States; to characterize the spatial distribution of these features within the area; and to present geospatial datasets of these features and their watersheds. The study area encompasses the Cumberland, Tennessee, Barren, and Conasauga River watersheds and eight karst regions with boundaries and names matching those of Level III and IV ecoregions data (U.S. Environmental Protection Agency, 2013): 1) the Inner Nashville Basin, 2) the Outer Nashville Basin, 3) the Eastern Highland Rim, 4) the Western Highland Rim, 5) the Western Pennyroyal Karst Plain and Crawford-Mammoth Cave Uplands, 6) the Cumberland Plateau Escarpment and Sequatchie Valley, 7) the Ridge and Valley, and 8) the Limestone Valleys and Coves. Digital elevation data were obtained from two sources: 1) the National Elevation Dataset (NED; Gesch and others, 2002; Gesch, 2007; U.S. Geological Survey, 2011 - 2012, variously dated a), which was derived from topographic-contour source data for some areas and light detection and ranging (lidar) source data for other areas, and 2) photogrammetrically-derived data collected by the State of Tennessee as part of the Tennessee State Base Mapping Program (TNBMP; Tennessee Department of Finance and Administration, 2005, 2007). These sources of elevation data, along with flowlines from the National Hydrography Dataset (NHD; U.S. Geological Survey, variously dated b), were used to delineate depressions from each elevation data source. The NED is available for the entire study area, making it possible to delineate depressions derived from the NED across the entire study area. The TNBMP elevation data covers only parts of the study area so the depressions were delineated for parts of Tennessee. Many depressions in DEMs can be considered artificial due to errors in the source DEM (Lindsay and Creed, 2006; Zandbergen, 2010) or misrepresentation of surface water flow beneath elevated structures present in the DEM (Poppenga and others, 2010; Zandbergen, 2010; Wall and others, 2015). While many of the depressions in DEMs are artificial, some are real, such as those representing sinkholes found in a karst landscape. Initially, preliminary depressions, which include depressions classified as either likely or unlikely to exist on the landscape, were delineated from each elevation source. Assessment methods, including an estimation of elevation data accuracy, a numerical error propagation analysis of the elevation data in test areas, and a comparison of preliminary depression locations to topographic-contour source data in middle Tennessee and northern Alabama, were used to determine preliminary depression characteristics and identify proper depression-characteristic thresholds for distinguishing likely from unlikely depressions in both the NED and the TNBMP elevation data. Depressions included in this data release were those that 1) passed proximity filters, and 2) met the depression-characteristic thresholds. See the "Delineation of Depressions" processing step in the metadata document for more information about depression requirements. The DEM analysis methods used for this study provide no distinction between sinkholes, man-made depressions, or natural depressions; such distinctions are beyond the scope of the study and this data release. Depressions in the geospatial datasets represent areas of internal drainage that have the potential to store surface runoff. A given depression within the dataset may contain one, more than one, or no sinkholes depicted on a 1:24,000-scale topographic map. Flowlines from the NHD were examined and edited to produce a dataset of sinking streams within the study area, and watersheds for likely depressions and sinking streams were delineated from the NED and the TNBMP elevation data. During the NHD editing process, long flowline connectors and other NHDFlowline features (U.S. Geological Survey, variously dated b) that are not representative of natural surface-water flow paths were removed where appropriate to create a subset of sinking streams represented by individual NHDFlowline features and groups of connected NHDFlowline features that conceptually lose all surface-water flow to the subsurface at a termination point. In many cases, these sinking streams terminated within a sinkhole depicted on a 1:24,000-scale topographic map. If one or more sinking streams terminated within a likely depression identified in a DEM, the associated watershed contained the depression and all sinking streams flowing into it. If a sinking stream did not terminate within a likely depression, the associated watershed contained the sinking stream termination location and all stream lines upstream of the termination point. This collection of geospatial data includes 5 shapefiles, each organized into individual zip files: (1) depressions derived from the NED (NED_depressions.zip); (2) watersheds for depressions derived from the NED and sinking streams (NED_watersheds.zip); (3) depressions derived from the TNBMP elevation data (TNBMP_depressions.zip); (4) watersheds for depressions derived from the TNBMP elevation data and sinking streams (TNBMP_watersheds.zip); and (5) sinking streams (sinking_streams.zip).