mineBenchDL: A geomorphology deep learning dataset of historic surface coal mine benches in West Virginia, USA

mineBenchDL DatasetA dataset for geomorphic deep learning. Goal is to detect historic mine benches resulting from surface coal mining from lidar-derived digital terrain models. Mine bench extents were manually digitized via interpretation of lidar-derived terrain data, leaf-off aerial imagery, and ancillary geospatial data. Mine bench extents have been provided as vector data. The elev folder contains the lidar-derived elevation data at a 2 m spatial resolution. The images folder contains a 3 band stack of terrain variables. The masks folder contains a binary feature mask. Data were randomly split into training, test, and validation sets randomly using quarter quad boundaries*All layers in NAD83 UTM Zone 17NtrainQQs.csv: quarter quads randomly selected as training set (69 quarter quads)testQQs.csv: quarter quads randomly selected as test set (49 quarter quads)valQQs.csv: quarter quads randomly selected as validation set (50 quarter quads)vectors folder: mineBenches.shp = digitized mine bench features (QKEY field used as unique ID) quarterQuads.shp = quarter quad boundaries (just those containing mine benches)images folder: 3 band stack of land surface parameter predictor variables (2 m spatial resolution). The first layer is a topographic position index (TPI) calculated using a moving window with a 50 m circular radius and designed to characterize general hillslope position. The second layer is the square root of slope calculated in degrees, which provides a measure of steepness. The third layer is another TPI; however, it is calculated using an annulus moving window with an inner radius of 2 and outer radius of 5 m. This surface captures more local relief and surface roughness patterns in comparison to the other TPI. Values for both TPIs and the square root of slope were clamped to a range of -10 to 10 then linearly rescaled to a range of 0 to 1. masks folder: Mine bench raster masks (1 band, 1 = bench, 0 = background)<br>