Predicting the Probability of Concentrated Flow Occurrence on Rangeland with Three-dimensional Data

Abstract. In hillslope erosion modeling, concentrated flow pathways are represented as fluvial processes distinct in hydraulics and hydrologic behavior from areas where splash and sheet flow dominate. As a result, many physically-based hillslope erosion models distinguish between splash and sheet (or interrill) erosion processes from concentrated flow processes also called rill erosion. While the theories and equations governing transport and erosion in concentrated flow channels have been well studied, more research is needed to better predict the presence of channels on rangeland. As opposed to cropland where rills can be assumed to follow tillage marks, no a-priori assumption can be made on the occurrence and distribution of concentrated flow pathways on rangeland. Rain splash and sheet erosion processes are typically the dominant erosion processes on undisturbed rangelands, resulting in low erosion rates compared to levels observed on cropland. On disturbed rangelands, loss of vegetation, changes to soil properties (e.g., hydrophobicity after fire) often lead to increased risk of concentrated flow development and accelerated erosion rates (Pierson et al., 2011; Williams et al., 2014). Using experimental data from concentrated flow runoff experiments, Al-Hamdan et al. (2013) proposed the first model to predict the likelihood that a flow released over a rangeland area will become concentrated based on flow and surface characteristics. Flow concentration on rangeland occurs as runoff from interrill areas either concentrate into pre-existing channels or form new channels on initially undissected hillslopes. These two flow concentration situations have different hydraulic implications but are undistinguished in current soil erosion modeling schemes. The aim of this study is to propose a predictive model for the existence of flow concentration pathways on rangeland soil surfaces based on soil, topography, and vegetation characteristics. The study uses three-dimensional data collected during rainfall simulation experiments on 6 m x 2 m plots across multiple rangeland sites of the western United States. Three-dimensional data consist in pre- and post- rainfall simulation digital elevation models (Figure 1) obtained using Structure-from-Motion surface reconstruction. Using geomorphological concepts and methods, the channel network associated with a given soil surface can be extracted (Figure 1A) and relationships established between the likelihood of channel presence and soil, plants, and topographic attributes. New knowledge from this study will improve rangeland erosion models such as RHEM.