Numerical ocean models are now being applied in numerous oceanographic stud
ies. However, the qualities of the model results are often uncertain and th
ere is a great need for standards and procedures for evaluation of the skil
ls of numerical general circulation models. In this paper measurements from
repeated hydrographical sections across Skagerrak taken in 1990, the SKAGE
X dataset, are used to evaluate the skills of two cr-coordinate ocean model
s and to study the sensitivity of these models to model parameters. A metho
dology for quantification of model skills based on observations from repeat
ed hydrographical sections in general is suggested. Area averages of absolu
te differences are for Skagerrak completely dominated by the discrepancies
in the upper few meters of the ocean and may not be used to assess models'
abilities to reproduce the fields in the larger and deeper part of the ocea
n. Therefore, discrepancies between average values in time from the observe
d fields and time averaged values from model outputs are related to the nat
ural variability of the fields. The numbers produced with the suggested mea
sure are relative numbers that will be specific for each section and for ea
ch series of observation. Ideally we would therefore like to see the measur
es computed for a number of sections for various models and choices of mode
l parameters in order to assess model skills. The value of the SKAGEX datas
et as a tool for model improvements is demonstrated. Evidence to support th
e importance of applying non-oscillatory, gradient preserving advection sch
emes in areas with sharp density fronts is given. The method is used to ide
ntify that the forcing/initial values/boundary values for the temperature f
ield are inferior to the corresponding values for the salinity field. With
the present coarse resolution, 11 layers in the vertical, it is shown that
it is far from obvious that the quality of the model results improve when r
eplacing simple Richardson number formulations for vertical mixing processe
s with higher order turbulence closure in the Skagerrak. area. (C) 1999 Els
evier Science B.V. All rights reserved.