Internal pressure gradient errors in σ-coordinate ocean models in high resolution fjord studies

Terrain following ocean models are today applied in coastal areas and fjords where the topography may be very steep. Recent advances in high performance computing facilitate model studies with very high spatial resolution. In general, numerical discretization errors tend to zero with the grid size. However, in fjords and near the coast the slopes may be very steep, and the internal pressure gradient errors associated with σ-models may be significant even in high resolution studies. The internal pressure gradient errors are due to errors when estimating the density gradients in σ-models, and these errors are investigated for two idealized test cases and for the Hardanger fjord in Norway. The methods considered are the standard second order method and a recently proposed method that is balanced such that the density gradients are zero for the case ρ = ρ ( z ) where ρ is the density and z is the vertical coordinate. The results show that by using the balanced method, the errors may be reduced considerably also for slope parameters larger than the maximum suggested value of 0.2. For the Hardanger fjord case initialized with ρ = ρ ( z ) , the errors in the results produced with the balanced method are orders of magnitude smaller than the corresponding errors in the results produced with the second order method.