We have had much greater difficulty in replicating the results with the NCEP pressure archive than from ECMWF, and differences were well in excess of 100 W m-2. The main problem is pathological in the NCEP reanalyses in the stratosphere. It is manifested most strongly as a two-delta vertical wave in the divergence of the wind field above steep topography especially where the wind increases with altitude. It is present primarily above 50 mb at the topmost four levels in the NCEP model used for data assimilation and appears to be directly related to the use of the sigma (terrain-following) coordinate system and the upper boundary condition in the assimilating model. Recommendations are suggested for addressing the problem.
We also address the question of how well model level diagnostics can be replicated with a coarser resolution pressure-level archive and recommend that future pressure archives should have higher vertical resolution to better resolve the boundary layer and the surface.