If it is 1D, I would increase the roughness coefficient of the US and DS cross sections. If it is 2D I would also adjust the associated cells roughness coefficient assuming they were small enough to not impact adjacent cells.
Another option could be to model the weir as a cross section with Lids. In this instance you could correlate an exact cross section with a weir and associated roughness coefficient. You could also add a minor loss coefficient to the model in the Table of the geometry data editor.