Exploring the effect of assimilating BGC-Argo observations on the Met Office's marine biogeochemical forecasting system

The UK Met Office produces operational forecasts for the Northwest European Shelf region, for which we routinely assimilate temperature and salinity from Argo but not yet biogeochemical (BGC)-Argo variables. To explore the effect, we conduct a set of data assimilation experiments with measured and machine learning-derived BGC-Argo data. We use the NEMOVAR assimilation scheme in its 3DVar configuration with first guess at appropriate time. Biogeochemical variables are assimilated univariately, meaning that in the assimilation step each variable is updated individually and the model dynamics distribute the changes to other model variables.

The first attempt of assimilating nitrate revealed a bias which we traced to the lateral boundary conditions, which will shortly be fixed in the operational system. A question for assessing the benefit of assimilating BGC-Argo data is how widespread the effect is of assimilating the relatively sparsely distributed BGC-Argo profiles, which only cover the off-shelf area. We find that the spread of the assimilated information depends on the interior circulation, which is affected by the assimilation of physics variables, including how much of the signal is advected onto the shelf. Assimilating nitrate from BGC-Argo also had effects on non-assimilated variables such as the distribution of chlorophyll within and below the mixed layer, and it revealed mismatches in the vertical structure of observed nitrate and the model’s mixed layer depth. The results suggest that our forecasting system can benefit from assimilating BGC-Argo data, directly through the assimilation as well as indirectly by highlighting issues in the physics-BGC interactions. Future work should investigate how to better match the assimilated physics, and BGC and also explore balancing the assimilated information from each variable across the ecosystem to increase the impact of the observations.