ASABE Technical Library - Abstract

Abstract.

The precise determination of thresholds over which topographic index (TI) models perform better require a proper understanding of ephemeral gully (EG) locality and characteristics. This study aims at evaluating the accuracy of six TI models at prediction of ephemeral gullies (EGs) within Running Turkey watershed in south-central Kansas. TI models applied include: slope area model (SA), compound topographic index model (CTI), wetness topographic index model (WTI), slope area power (SA2), kinematic wave model (nLS), and modified kinematic wave model (nLSCSS). EGs predicted by each model threshold were compared with actual EGs obtained through digitization and field reconnaissance or “ground truthing”. Statistical analysis was performed using the approach of confusion matrix for location analysis, and root mean square error (RMSE) and Nash–Sutcliffe efficiency (NSE) for length analysis. TI models were able to predict EG location with a maximum accuracy of 70%. The CTI (kappa = 0.29, NSE =0.8) model outperformed all the models at prediction of EGs. The nLS and nLSCSS models had poor performance at prediction of EGs, which is attributed to inaccuracies at precise computation of Manning’s and critical shear stress values at each pixel within catchments. TI models can predict EG location and length over a range of thresholds rather than one single value. The accuracy and range of TI thresholds require using a holistic approach of considering both EG location and length to evaluate the optimum threshold values over which TI models predict EGs better.