Just as a comment, I’m wondering whether the loss coefficients of 0.3 et 0.5 which are applied in a 1D model should also be applied in a 2D model. I’d say not, as some of the losses are already included in the 2D model, but I’m not 100% sure.
One thing that wasn’t obvious to me when I modelled my first bridge as a culvert with RAS2D was that I didn’t know which cells were considered for the culvert hydraulics calculation. I’ve found out there are only two cells even if the 2D area connector spans over many more: one upstream and one downtream. This is different from the weir flow calculation, which will take into account all the cells adjacent to the 2D area connector. Because only two cells are considered for culvert flow calculation, you should make sure to select a small time step or adapt the size of these two cells so that too much flow cannot travel from one cell to the other over a single time step. It’s easy to picture that things would go really wrong if the solver tries to compute the flow between two really tiny cells (say 0.1 * 0.1 m) based on a very large culvert (say 20m wide by 5m high). You’d need a super small timestep not to empty the upstream cell and not to generate super high flow depth at the downstream cell, otherwise there would obviously be some oscillations.
One trick to preserve small cells and a decent time step could be to model the bridge opening(s) with many culverts whose culmulated areas match with the bridge opening(s) area. I haven’t tested that out but I guess that it could work, in theory. The loss coefficients should certainly be adapted not to cumulate them.