Even though the basic problem of the terrain-following vertical coordinates has been in a step-by-step fashion revealed, for the most part, in a sequence of papers by Rousseau and Pham (1971), Janjic (1977), Mesinger (1982), and Mesinger and Janjic (1985), the problem appears to have received little attention by the NWP community. In numerous ambitious model-development projects the issue has been ignored, in spite of the widespread movement toward higher resolution, which should make the resulting errors more severe. The nature of the problem is summarized, aiming for a statement more general than the one of Mesinger and Janjic (1985).

One model that radically circumvents the problem is the Eta model, using quasi-horizontal coordinates. An update on the performance of the Eta at NCEP is presented, including precipitation scores, rms fits to raobs as a function of time and season, and accuracy in placement of centers of major storms. In a number of aspects the verification results reported, viewed in comparison with the performance results of NCEP's terrain-following coordinate models, are consistent with and suggestive of the possibility of the eta coordinate making a significant contribution to the favorable performance of the Eta. One of these are precipitation scores for the very heavy rainfall over the U.S. high topography western region during the late fall 2002 and winter 2002-2003, with the Eta 12 km model considerably outperforming an 8 km resolution terrain-following coordinate model. There is on the other hand little difference in the scores of the two models in the eastern U.S. region, not having a pronounced topography.

Given that with typical numerical formulations what one is facing with increasing realism of topography and thus increased steepness of coordinate surfaces is loss of information needed to calculate the slopes of pressure surfaces, it is suggested that this issue may well be one deserving much more attention that it presently does.