Abstract
A native nested configuration of the ROMS model is implemented on the marine area between the Ligurian and Tyrrhenian basins, which includes the Tuscany Archipelago. Initial and boundary conditions are provided by the CMEMS Mediterranean Sea Physical Reanalysis product (1/16°), feeding the parent ROMS model (BLUE, 1/72°), in which a high-resolution grid is nested (PURPLE, 1/216°). Atmospheric forcing comes from a downscaled version of ERA5 reanalysis. Temperature and salinity profiles from gliders and floats, and HF-radar-derived surface currents, are compared to model outputs within the high-resolution area for the whole year 2017. Results show the downscaling procedure is able to reduce model errors for temperature profiles, whereas errors in salinity profiles remain comparable. However, the downscaled model cannot recover large errors inherited from the parent one. The mean bias largest values found in both temperature and salinity profiles may be explained by a model underestimation of the depth of stable stratification limit with respect to field data. Errors in surface currents are reduced for the downscaled dynamics and appear to be uncorrelated to the original CMEMS product, being surface dynamics less affected by initial condition than by atmospheric forcing. A simple scalar metric, to quantify the error in the surface current vector fields from observations and models, is proposed. The novel metric allows to better quantify the improvement gained by the downscaling procedure.