Abstract
Traditionally, in order for lower-resolution, global- or basin-scale (regional) models to benefit from some of the improvements available in higher-resolution subregional or coastal models, two-way nesting has to be used. This implies that the parent and child models have to be run together and there is an online exchange of information between both models. This approach is often impossible in operational systems where different model codes are run by different institutions, often in different countries. Therefore, in practice, these systems use one-way nesting with data transfer only from the parent model to the child models. In this article, it is examined whether it is possible to replace the missing feedback (coming from the child model) by data assimilation, avoiding the need to run the models simultaneously. Selected variables from the high-resolution simulation will be used as pseudo-observations and assimilated into the low-resolution models. This method will be called “upscaling”. A realistic test case is set up with a model covering the Mediterranean Sea, and a nested model covering its north-western basin. Under the hypothesis that the nested model has better prediction skills than the parent model, the upscaling method is implemented. Two simulations of the parent model are then compared: the case of one-way nesting (or a stand-alone model) and a simulation using the upscaling technique on the temperature and salinity variables. It is shown that the representation of some processes, such as the Rhône River plume, is strongly improved in the upscaled model compared to the stand-alone model.