Influence of Ephemeral Gully Location Prediction on Soil Erosion Estimation

Predicting ephemeral gully (EG) location is essential for erosion modeling because it helps confine portions of the hillslope segment above locations that gully and channel soil loss processes dominate. In the Water Erosion Prediction Project (WEPP), the prediction of EG occurrence location influences the model results by shorting or expanding the flow path, which the hillslope erosion modeling relies on. This research aimed to analyze the sensitivity of EG locations prediction accuracy on WEPP model output within the framework of the Daily Erosion Project (DEP) at the regional scale. DEP is a near real-time estimator of precipitation, soil detachment, hillslope soil loss, and water runoff using WEPP as the erosion model. The above estimations are conducted on randomly selected and spatially distributed flowpaths, and the means are reported at the HUC12 watershed level. The flowpaths are identified based on Digital Elevation Model (DEM) grid cell and D8 connectivity to adjacent cells. A flow path starts at a cell such that all adjacent cells are at a lower elevation, that is, no other adjacent cell directs flow into it and ends when sufficient flow concentration and soil conditions occur that channel erosion processes dominate soil loss where usually EGs occurrence. In this research, the DEP flowpaths, down to and including ephemeral gully heads, were surveyed in 8 HUC12 watersheds distributed in 8 different Iowa MLRAs using high-resolution imagery in-field measurement. A grid order model was used as a method for EG location prediction. The sensitivity of accuracy of EG location prediction on WEPP/DEP soil detachment, hillslope soil loss, and water runoff model output was explored at hillslope, watershed, and regional spatial scale with both extreme rainfall events and yearly average erosion modeling. This research will allow a more clear understanding of EG prediction influence on erosion modeling and help improve the accuracy of erosion modeling by using WEPP / DEP.