SUMMARY
The prediction of water levels in rivers is important to prevent economical as well as human losses caused by flooding. Hydraulic models are commonly used to predict those water levels and take actions to mitigate flooding damage. In this research, a 2D approach to solve the depth average Reynolds Average Navier Stokes (RANS) equations, called Conveyance Estimation System (CES), is analyzed to explore its capabilities for prediction. This article presents an extension of the study performed in Knight et al. (2009). More specifically, in this study, a more detailed characterization of the roughness parameter and the number of roughness zones is explored producing additional scenarios. The performance of each scenario is evaluated by means of different fitting functions using rating curves for comparison. The research shows that the use of an adequate roughness description, such as a roughness factor calibrated for the whole cross section or a boulder roughness model calibrated for the channel bed plus roughness values from the CES roughness advisor for banks, leads to optimal model results in a mountain river.