We observed the incisional response of an alluvial river to base-level fall. We conducted the experiment in a 3.9 × 2.4 × 0.4 m box that we filled with uniform 0.140±0.04 mm sand. We dropped base level by lowering the elevation of an "ocean" pool at the river outlet. As the initial condition, we cut a 10±2 cm wide channel to a steadily increasing depth, from 3±0.5 cm at the inlet, where we supplied water and sediment, to 10±1 cm at the outlet. Input water and sediment discharge were 0.1 L/s and 0.0022 L/s (including pore space), respectively. As base level fell, the river incised and migrated laterally, forming a valley with abandoned terrace surfaces and walls that failed in mass-wasting events as they were undercut. We include a control case with no base-level fall, as well as experiments with 25 mm/hr, 50 mm/hr, 200 mm/hr, 300 mm/hr, and 400 mm/hr of base-level fall. We supply georeferenced overhead photos (0.89 mm resolution, every 20 seconds), digital elevation models (DEMs, 1 mm horizontal resolution, every 15–30 minutes), videos generated from the overhead photos, mapped landslides in GIS vector area (polygon) format, and landslide attributes. Relevant code to process and plot the data, as well as further information on grain size, is available from GitHub and Zenodo.
The data set contains the following items. All files with time-dependent data contain 7-digit time stamps that provide the number of seconds since the start of each experiment.
- Georeferenced overhead photos in GeoTIFF format within a Tape ARchive (tar).
- Digital elevation models in GeoTIFF format within a GZipped Tape ARchive (tar.gz).
- Landslide GIS vector areas (polygons) organized by experiment and time as ESRI Shapefiles stored within a GZipped Tape ARchive (tar.gz). Each subfolder is labeled <TIMESTAMP>.shx (the ".shx" is spurious) and contains the shp, shx, dbf, and prj files associated with each set of landslides at each time in the experiment.
- Landslide attributes as comma-separated variables (csv) files for each experiments, stored within an XZipped Tape ARchive (tar.xz). Each CSV includes (in order): the x and y positions of the landslide center, the width (y-directed – i.e., cross-valley – distance from one end to the other) and length (x-directed – i.e., down-valley – distance from one end to the other) of the landslide, mean landslide depth, landslide area measured in the x-y (i.e., horizontal) plane, the semi-major and semi-minor axes of an ellipse fit to the landslide, the angle from the semi-major axis to the "horizontal" (confusingly meaning the x orientation, since I was thinking in x-y space while writing the analysis code), landslide volume calculated by subtracting the valley-bottom elevation from that of the DEM surface in the landslide area, the runtime at which the landslide occurs, and the wait time between landslide events.
- Movies at 15 fps (5 minutes experiment time per 1 second watch time) for the full series of images from each experiment, stored within an XZipped Tape ARchive (tar.xz). I did not adjust for occasional skipped images (e.g., around the time of the laser scans), which will cause minor deviations from the 5-minutes-runtime-to-1-second-video conversion.
- Schematic image (png) of a georeferenced overhead photo (ImgSec_0043168) atop a shaded-relief map (DEM_fullextent_0043188) with hatched landslide locations from runtimes after 0043168. Shadows running along the x axis show zones near the outlet where the basin walls prevented the angled laser-topography scanner from casting light on the the valley bottom. All images are from the 25 mm/hr base-level fall experiment.