Erosion rates and salinity and selenium yields in a basin near Rangely, Colorado following the 2017 Dead Dog wildfire as modeled by WEPP and measured from UAV

This data release accompanies a U.S Geological Survey study that assessed sediment, salinity, and selenium yields following the Dead Dog wildfire in northwestern Colorado. The Dead Dog fire ignited on June 11, 2017, near Rangely, Colorado, and burned over 17,000 acres, including the B2 study area. Two methodologies were used to quantify erosion and associated salinity and sediment yields in the B2 study area: (1) modeled soil loss post-fire using a physically based erosion model, and (2) measured post-fire volumetric soil erosion and deposition using a time series of digital elevation models (DEMs) created from Uncrewed Aerial Vehicle (UAV) imagery.

The first methodology used a physically based erosion model, the Watershed Erosion Prediction Project (WEPPcloud; WEPP hereafter). The WEPP model used inputs of climate, topography, vegetation, and soils data from existing datasets to predict erosion. Three model runs were completed with WEPP. First, erosion was modeled for the year prior to the fire (2016). This model run simulated pre-fire erosion and provided an initial groundwater storage value used in subsequent model runs. Second, erosion was modeled for 2017 to 2021 under unburned conditions. Third, erosion was modeled for 2017 to 2021 under burned conditions. The burned model run used the same basin information (reaches, hillslopes, and catchments), slope information, and climate information as the 2017 to 2021 unburned scenario but also included a burn severity raster. The burn severity raster is used by WEPP to modify the soils and vegetation to represent post-fire conditions.

The second methodology utilized data from pre-fire UAV flights previously flown over the B2 study basin on September 1, 2016, and data from post-fire UAV flights on October 5 and 7, 2021 to examine post-fire morphological changes. Specifically, orthomosaics and digital elevation models (DEMs) were created from UAV images from each set of flights with elevational changes (erosion or deposition) within each pixel calculated from a DEM of difference (DoD) between flights. We evaluated two approaches for aligning UAV images from each set of flights. In the first approach, the 2016 and 2021 imagery were aligned independently using separate ground control points placed prior to each set of flights. For the second approach, the 2016 and 2021 images were co-aligned using natural ground control points, located on invariant objects, that were identified from the independently aligned imagery and associated DEMs. Internal consistency of the DEMs created from both approaches were determined by comparing elevation values at 100 check points on invariant surfaces located across the basin. The DEMs from the co-aligned imagery had greater internal consistency, with errors being normally distributed and a lower mean error compared to non-normally distributed errors and a greater mean error in the independently aligned imagery.
Therefore, a significant change DoD was created using elevational changes observed between the DEMs from the co-aligned imagery. First, non-bare ground pixels in the 2016 pixels were masked. Second, minimum detection limits were calculated using three standard deviations from the mean check point error. Finally, unrealistically high values associated with misclassified vegetation and along the escarpment rim and around large talus blocks were removed by using three standard deviations from mean of elevation changes outside the minimum detection limit. The significant change DoD was used, along with soil survey data, to calculate volumetric changes, erosion rates, and salinity and selenium yields across the entire basin (landscape extent) and within large stream channels below the escarpment rim (channel extent).

A complete list of files included in this data release is presented in the Overview section