Evaluating ephemeral gullies with a process-based topographic index model

Abstract

Soil conservation practices have been implemented to control soil degradation from sheet and rill erosion, but excessive sediment runoff remains among the most prevalent water quality problems in the world. Ephemeral gully (EG) erosion has been recognized as a major source of sediment in agricultural watersheds; thus, predicting location and length of EGs is important to assess sediment contribution from EG erosion. Geomorphological models are based on topographic information and ignore other important factors such as precipitation, soil, topography, and land use/land management practices, whereas physically based models are complex, require detailed input information, and are difficult to apply to larger areas. In this study, an approach was developed to incorporate a process-based Overland Flow-Turbulent (OFT) EG model that contained factors accounting for drainage area, surface roughness, slope, soil critical shear stress, and surface runoff in the ArcGIS environment. Two hydrologic models, Soil Water Assessment Tool (SWAT) and ArcCN-Runoff (ACR), were adopted to simulate precipitation excess in Goose Creek watershed in central Kansas, USA. These two realizations of the OFT model were compared with the Slope-Area (SA) topographic index model for accuracy of EG location identification and length calculation. The critical threshold index in the SA model was calibrated in a single field in the watershed prior to EG identification whereas the OFT models were uncalibrated. Results demonstrated overall similar performance between calibrated SA model and uncalibrated OFT-SWAT model, and both outperformed the uncalibrated OFT-ACR model. In simulation of EG location, the OFT-SWAT model resulted in 12% fewer false negatives but 8% more false positives than the SA model, compared with 19% fewer false positive and 6% more false negatives than the OFT-ACR model. Greater errors in runoff estimation by ACR translated directly into errors in EG simulation. All models over-predicted EG lengths compared with observed data, though OFT-SWAT and SA models did so with better fit exceedance probability curves, about zero Nash-Sutcliff model efficiency and ≤40% bias compared to -3 model efficiency and >100% bias for OFT-ACR. Success of the uncalibrated OFT-SWAT model in producing satisfactory predictions of EG location and EG length shows promise for process-based EG simulation. The OFT-SWAT model used data and parameters also commonly used for SWAT model development, which should simplify its adoption to other watersheds and regions. Further testing is needed to determine the robustness of the OFT-SWAT model to dissimilar field and hydrologic conditions. It is expected that inclusion of more site-specific physical properties in OFT-SWAT would improve model performance in predicting location and length of EGs, which is essential for accurate estimation of EG sediment erosion rates.