A framework for testing large-scale distributed soil erosion and sediment delivery models: Dealing with uncertainty in models and the observational data

Evaluating distributed soil erosion models is challenging because of the uncertainty in models and measurements of system responses. Here, we present an approach to evaluate soil erosion and sediment delivery models, which incorporates sediment source fingerprinting and sediment-rating curve uncertainty into model testing. We applied the Generalized Likelihood Uncertainty Estimation (GLUE) methodology to the Sediment Delivery Distributed model (SEDD) for a large catchment in Southeast Brazil. The model was not rejected, as 23.4% of model realizations were considered behavioral. Fingerprinting results and SEDD simulations showed a partial agreement regarding the identification of the main sediment sources in the catchment. However, grid-based estimates of soil erosion and sediment delivery rates were highly uncertain, which restricted the model?s usefulness for quantifying sediment dynamics. Although our results are case-specific, similar levels of error might be expected in erosion models elsewhere. The representation of such errors should be standard practice.

Automated summary

The study presents an approach to evaluate soil erosion and sediment delivery models by incorporating sediment source fingerprinting and sediment-rating curve uncertainty. GLUE methodology applied to SEDD model revealed partial agreement between fingerprinting results and simulations, but high uncertainty in grid-based estimates of soil erosion and sediment delivery rates. Such errors should be considered standard practice in erosion models.