Geomorphic responses to post-grazing recovery and stream restoration in semiarid grassland streams
Data files
Feb 26, 2026 version files 25.79 KB
-
README.md
5.47 KB
-
remote_sensing_bsi_clean.csv
2.38 KB
-
soil_bulk_density_clean.csv
8.64 KB
-
structures_basin_characteristics_clean.csv
9.30 KB
Abstract
Semiarid grassland streams are sensitive to land use, climate, extreme discharges, and internal geomorphic thresholds that drive episodic erosion. Rooted in a process-based philosophy and commonly applied to historically wood-rich, beaver-modified systems, low-tech process-based restoration using structures is increasingly being extended to other geomorphic settings, including grasslands. In these wood-poor systems with fine-grained substrate, channel adjustment depends primarily on subsurface soil piping and asynchronous responses to changing boundary conditions. We evaluate post-grazing recovery and structural restoration outcomes at transect-to-landscape scales in southeastern Colorado. Following grazing cessation, soil compaction decreased after roughly a decade of rest, as indicated by a multivariate linear regression analysis of 238 bulk density samples. In contrast, visible above-ground upland vegetation recovers on timescales longer than the observation window (>10 years), based on analyses of growing-season composite Landsat imagery from 2013–2024. Logistic regression models indicate that the performance of restoration structures depends strongly on drainage area, slope, local vegetation cover, and the recency of large-scale commercial grazing. In total, only ~30% of the 92 in-channel structures evaluated (mostly ≤4 years old) were both intact and produced visible geomorphic change. Structures in piping-dominated headwaters remained intact but rarely trapped sediment, whereas those in larger catchments were more likely to initiate local geomorphic change, despite having higher odds of destruction during peak flows. These findings highlight the need to match restoration strategies with the geomorphic processes driving channel adjustment and call into question the extent to which commonly implemented structure-based interventions in grassland systems are truly process-based.
Dataset DOI: 10.5061/dryad.4b8gthtsp
Description of the data and file structure
This dataset includes the code and processed data required to replicate analyses in Richardson and Wohl (2026), evaluating post-grazing recovery and in-channel stream restoration outcomes in semiarid grassland systems. Field data were collected in two small watersheds in southeastern Colorado during summer 2025. Remote data were collected from 1m LiDAR-derived DEMs (raw data can be obtained from the USGS 3D Elevation Program (3DEP): https://apps.nationalmap.gov/downloader/), Landsat 8/9 imagery collected over 2014-2024 (raw data can be obtained from USGS EarthExplorer: https://earthexplorer.usgs.gov/), and the NRCS Web Soil Survey (https://websoilsurvey.nrcs.usda.gov/app/).
Files and variables
File: structures_basin_characteristics_clean.csv
Description: This file contains restoration structure outcomes and geomorphic variables associated with the restoration structure location.
Variables
- Preserve: A categorical variable, indicating the Southern Plains Land Trust preserve where the structure of interest is located.
- Type: A categorical variable, indicating the type of restoration structure. Categories include ORD (one-rock dam), PALS (post-assisted log structures), Zuni (Zuni bowl erosion control structure), Logs (log step fall structures), and BDA (beaver dam analog).
- Year: The year the structure was installed.
- Geomorphic Change?: A Y/N variable indicating whether the structure resulted in visible geomorphic change (e.g., erosion, deposition, avulsion), as recorded during summer 2025.
- Intact?: A Y/N variable indicating whether the structure was intact as of summer 2025.
- Bank angle (degrees): The bank angle (numeric) in the immediate vicinity of the structure, measured with a laser range finder from the break in slope on the bed to the top of the adjacent alluvial surface.
- Bank vegetation: A categorical variable, describing the bank vegetation cover in the immediate vicinity of the structure, with levels of bare (<10% cover), sparse (10%, 50%) and heavy (>50% cover).
- Headcut: A Y/N variable, indicating whether the structure was located within three channel widths of a headcut.
- Drainage area (sqkm): The contributing catchment area upstream of the structure, calculated from a 1m LiDAR-derived digital elevation model.
- K factor: The RUSLE K-factor, describing soil erodibility, from the NRCS Web Soil Survey.
- Percent sand: The percent sand of the underlying soil, from the NRCS Web Soil Survey.
- Local Slope: The local channel slope calculated 10m upstream of the restoration structure from a 1m LiDAR-derived DEM.
- Grazing: A categorical variable, indicating the last presence of large-scale commercial grazing, with levels of "historic" (2015 or earlier) or "recent" (2021 or later).
- Longitude: Location (WGS84).
- Latitude: Location (WGS84).
File: soil_bulk_density_clean.csv
Description: This file contains soil bulk density data, collected across distinct land use history parcels and soil textures within the Southern Plains Land Trust Heartland Ranch Preserve.
Variables
- Parcel: A categorical variable indicating the soil bulk density sampling location grazing history (A-F).
- Texture Class: A categorical variable indicating the percent sand texture class with 5 corresponding to (0, 20%], 4 to (20, 40%], 3 to (40, 60%] and 2 to (60, 80%].
- Sample: For each parcel-texture class combination, we sampled bulk density ten times. This variable indicates the sample number (1-10) within each parcel-texture class combination.
- Mass (g): the mass of the soil sample after oven drying.
- Bulk Density (g/mL): Computed bulk density.
- Latitude: Location (WGS84).
- Longitude: Location (WGS84).
File: remote_sensing_bsi_clean.csv
Description: This file contains the median Bare Soil Index (BSI) value for each parcel and year 2014-2024, extracted from growing season (May-Sept.) Landsat 8/9 composites generated in Google Earth Engine.
Variables
- Parcel: A categorical variable indicating the sampling location (A-F).
- Year: The year over which the growing season composite was compiled.
- Median BSI: The median BSI value for each parcel-year combination.
- Grazing: A categorical variable, indicating the last presence of large-scale commercial grazing, with levels of "recent" (2015 or earlier) or "historic" (2021 or later).
- Growing Season Precipitation (mm): Annual growing season (May-Sept.) precipitation, from the nearby "Lamar3" CoAgMet weather station.
Code/software
R and Google Earth Engine code used to generate data and perform statistical analyses are available at https://github.com/richardson-owen/SPLTProject.
Access information
Data was derived from the following sources:
- USGS 3D Elevation Program (3DEP): https://apps.nationalmap.gov/downloader/
- USGS EarthExplorer: https://earthexplorer.usgs.gov/
- NRCS Web Soil Survey: https://websoilsurvey.nrcs.usda.gov/app/
