Predicting ephemeral gully location and length using topographic index models

Abstract

Ephemeral gullies (EGs) are incised channels resulting from concentrated overland flow that often form in a similar location every year. These erosional features add to producers’ management efforts and costs. Locating EGs and predicting their length is crucial for estimating sediment load and planning conservation strategies. Since topography plays an important role in the formation of EGs, this study investigated the prediction of EG location and length in two agricultural areas (S1 and S2) in two different physiographic regions using four topographic index models: compound topographic index (CTI), slope area (SA), wetness topographic index (WTI), and slope area power (SAP). The impacts of digital elevation model (DEM) resolution, agricultural land use mask data source, and topographic model critical thresholds were also evaluated. Automated geospatial models were developed to locate and derive EG length. Results show that the SA model predicted EG occurrence and length better than other models tested. The SA and CTI model predictions had similar patterns in terms of locating EG trajectory; however, the CTI model had greater discontinuity along the trajectory. The method developed to derive length in this study was sensitive to discontinuity, so the performance of the CTI model was poor. Finer-resolution DEMs (2 m) predicted EG location and lengths better than coarser-resolution DEMs (10 m or greater). Use of actual field-level reconnaissance data instead of NASS data for agricultural land use masking decreased false negative classification by 16% or more for all models. Detailed calibration of the SA model yielded different optimal thresholds for the two study regions: T[subscript SA] = 30 for S1 and T[subscript SA] = 50 for S2. Topographic index models were found to be useful in locating EGs and estimating expected lengths, but site-specific calibration of the topographic index model threshold was required, which might limit the general utility of these methods.