Abstract:
Three alternative methods to remove oscillations in two-dimensional Eulerian-Lagrangian salt transport simulations are compared: a Flux-Corrected Transport (FCT) scheme, a non-linear filter (NLF) and the explicit introduction of diffusion. The three formulations are implemented in a high-accuracy Eulerian-Lagrangian model (ELM), and are applied to the salinity propagation in the Guadiana estuary (Portugal). This test allows both for a stringent comparison in a real setting where adequate field data is available, and to assess the practical consequences of the choice of the numerical scheme. The evaluation of the schemes confirms a previous analysis showing that the FCT-based method provides the best results, closely followed by the non-linear filter. For these methods, oscillations are removed with minimal numerical damping, and the agreement with field data is good. The performance of the base model with non-zero diffusion coefficients is case-dependent, improving its results as the mixing of the front increases due to larger river flows. For the Guadiana estuary, the choice of the numerical scheme is crucial: important indicators, such as the domain-integrated residence time and the limit of salinity intrusion, may have errors up to 25%, if an artificially diffusive method is used to avoid numerical oscillations.