Accounting for Conceptual Soil Erosion and Sediment Yield Modeling Uncertainty in the APEX Model Using Bayesian Model Averaging Academic Article uri icon


  • 2014 American Society of Civil Engineers. The effects of soil erosion and sedimentation are important for natural resources conservation planning. However, although tremendous resources have been invested in developing more erosion models, the prevailing modeling studies are relying on a single model due to various reasons. The Agricultural Policy Environmental Extender (APEX) provides multiple water erosion equations. This study tests and evaluates the Bayesian model averaging (BMA) scheme on sediment predictions based on four water erosion methods in the APEX model using data from two watersheds. The APEX hydrology and soil erosion and sedimentation components were calibrated simultaneously using the APEX autocalibration tool APEX-CUTE. The BMA scheme is employed to obtain consensus predictions by weighing individual predictions based on their probabilistic likelihood measures. Simulated monthly flow was satisfactory for both the calibration and validation periods, with BMA resulting in Nash-Sutcliffe efficiency (NSE) values from 0.56 to 0.87 and percent bias (PBIAS) within 17%. No individual soil erosion method in APEX is consistently outperformed other method over the entire sediment yield regimes at both watersheds. The BMA scheme led to improved predictions compared with individual methods and resulted in satisfactory performance statistics (NSE from 0.55 to 0.88) for both the calibration and validation periods based on the optimal solutions. BMA also led to relatively higher Brier scores and narrower prediction bounds than individual methods. The methodology is unrestrictedly general and can be applied to any other combination of models in the environmental sciences.

published proceedings

  • Journal of Hydrologic Engineering

author list (cited authors)

  • Wang, X., Yen, H., Jeong, J., & Williams, J. R.

citation count

  • 11

complete list of authors

  • Wang, X||Yen, H||Jeong, J||Williams, JR

publication date

  • November 2014