Dear authors,

Please reflect on the comments of Referee 2, reproduced below, and proceed with appropriate technical corrections.

Best Regards,
Bob Marsh

Anonymous Referee #2, comments on revised manuscript:
I would like to thank the authors for the additional information given in the answers. The manuscript is an interesting description of the model mathematical background and gives insights and possible solutions to common numerical problems of unstructured mesh models.
But still the results do not convince me with respect to the validation in the realistic cases of the Sylt-Romo and South-East North Sea circulation experiments. The results seem to depend stronger on the applied forcing (sensitivity to open boundary, river input, etc) than on the numerical mesh (see differences in Fig 7 compared to over- respectively underestimation of the tidal amplitude maybe due to the different open boundary forcing in the two experiments?). The velocity differences between the meshes do not exceed 5 cm/s (mostly less than 1 cm/s). So the question is, what is absolute velocity and the natural variability of the currents in these areas and how does this difference compare to velocity changes due to uncertainties in the forcing, parametrizations and bathymetry. Are there still improvements in the numerical schemes required? The salinity and temperature comparison with observation could be more detailed, because a RMSD of 1.24 °C is quite common for ocean models in this region. The gradients seem to be stronger than observed. So I recommend setting up the model with better meteorology and open boundary forcing (including a surge model), tune it and then perform a more detailed validation with observation. Especially the storm surge heights have not been analyzed at all. (The sea surface height average for one tidal period could be zero, although there is atmospheric influence, e.g. during permanent easterly winds lowering the sea level in the German Bight). The question is, what the models purpose is, so what are the important variables to analyse.
The article has improved with the corrections and I would suggest accepting the manuscript as is (maybe do typo corrections: p.9 l.2 trough -> through and p.16 l.5 A, phi_star -> A, phi). Thanks for making code available in open source.