Assessment of the Skill of Coupled Physical-Biogeochemical Models in the NW Mediterranean

Abstract

Numerical modeling is a key tool to complement the current physical and biogeochemical observational datasets. It is essential for understanding the role of oceanographic processes on marine food webs and producing climate change projections of variables affecting key ecosystem functions. In this work, we evaluate the horizontal and vertical patterns of four state-of-the-art coupled physical–biogeochemical models, three of them already published. Two of the models include data assimilation, physical and/or biological, and two do not. Simulations are compared to the most exhaustive dataset of in situ observations in the North Western Mediterranean, built ad hoc for this work, comprising gliders and conventional CTD surveys and complemented with satellite observations. The analyses are performed both in the whole domain and in four subregions (Catalan Shelf, Ebro Delta, Mallorca Channel, and Ibiza Channel), characterized by a priori divergent primary production dynamics and driving mechanisms. Overall, existing models offer a reasonable representation of physical processes including stratification, surface temperature, and surface currents, but it is shown that relatively small differences among them can lead to large differences in the response of biogeochemical variables. Our results show that all models are able to reproduce the main seasonal patterns of primary production both at the upper layer and at the deep chlorophyll maximum (DCM), as well as the differential behavior of the four subregions. However, there are significant discrepancies in the local variability of the intensity of the winter mixing, phytoplankton blooms, or the intensity and depth of the DCM. All model runs show markedly contrasting patterns of interannual phytoplankton biomass in all four subregions. This lack of robustness should dissuade end users from using them to fill gaps in time series observations without assessing their appropriateness. Finally, we discuss the usability of these models for different applications in marine ecology, including fishery oceanography