Improving marine biogeochemical forecasts through data assimilation of BGC-Argo float data

Gianpiero Cossarini, Anna Teruzzi, Stefano Salon and Laura Feudale, National Institute of Oceanography and Applied Geophysics (OGS), Oceanography, Trieste, Italy
Abstract:
The increased availability of NRT data from BGC-Argo floats, besides satellite surface chlorophyll, provides the opportunity to improve the forecast skill of marine biogeochemical forecasts.

In the framework of the EU Copernicus Marine Environment Monitoring Services (CMEMS) an operational system for the short-term forecast of Mediterranean Sea biogeochemistry already includes the assimilation of ocean color satellite observations. The existing 3D variational assimilation scheme accounts for the horizontal and vertical error covariances among with the biogeochemical error covariance. An update of the error covariance operators has been designed for the assimilation of nitrate and chlorophyll vertical profiles of BGC-Argo floats. In particular, new biogeochemical covariances are set variable in time and space to account for different processes driving biogeochemical dynamics. Further, different combinations of assimilated and updated variables have been tested, and novel skill performance metrics developed to evaluate the impact of the different sets of observations on the forecast skill.

Results show that BGC-Argo observations improve the forecast skill performances increasing the persistency of the chlorophyll and nutrients update. Furthermore, BGC-Argo floats data assimilation has local but very positive impacts on the vertical dynamics of phytoplankton and nutrients. In particular, dynamics and temporal evolution of peculiar zonal gradients of the Deep Chlorophyll Maximum and the nutricline in the Mediterranean Sea can be effectively reproduced and investigated by the upgraded forecast system constrained by BGC-Argo observations.

For joint float-satellite chlorophyll data assimilation, skill and persistency metrics are affected by some inconsistency between the two data streams, probably related to different spatial observation scales and different measurement techniques.