Ocean Dynamics (2019) 69:1217–1237 1219 (e.g. Nerger and Gregg 2007, 2008; Ciavatta et al. 2011; Pradhan et al. 2019) or optimal interpolation (Ford et al. 2012) have applied the data assimilation to the logarithm of the concentrations or by applying a so-called anamorphosis transformation (Doron et al. 2011). For the BGC assimilation with variational methods, Song et al. (2016c) have developed a method to treat log-normal concentration distributions. On the other hand, the actual concentrations have been used by other studies applying ensemble Kalman filters (e.g. Carmillet et al. 2001; Natvik and Evensen 2003; Mattern et al. 2010; Yu et al. 2018) and 3-dimensional variational assimilation (Teruzzi et al. 2014). The latter study also discussed that actual concentrations were used because only then the typical structure of vertical chlorophyll profiles was preserved. In this study, both cases of actual and logarithmic concentrations are examined. This study is structured as follows: Section 2 describes the coupled model HBM-ERGOM. The data assimilation methodology and the observations assimilated and used for validation are described in Section 3 while Section 4 describes the setup of the data assimilation experiments. The assimilation effect is assessed in Section 5 for using actual biogeochemical concentrations and in Section 6 for the case of the logarithmic treatment of the biogeochemical variables. The results are discussed in Section 7 while conclusions are drawn in Section 8. 2 HBM-ERGOMmodel The model used here is the HIROMB-BOOS-model (HBM) coupled to the BGC model ERGOM. HBM is currently used operationally, without data assimilation, by the BSH in a similar configuration as used here. The coupled HBM- ERGOM configuration is currently used pre-operationally at the BSH. HBM is a three-dimensional hydrostatic circulation model using the primitive equations. It uses spherical horizontal and generalised vertical coordinates (Kleine 2003). The model domain extends from 4? W to 30.5? E and from 48.5? N to 60.5? N in the North Sea and to 66? N in the Baltic Sea. A nested configuration of the model is used with two domains shown in Fig. 1. The coarser grid covers the entire North Sea and Baltic Sea. It has horizontal grid spacing of about 5 km (5’ in longitude and 3’ in latitude) and 36 vertical layers. In the region of German territorial waters in the North Sea and Baltic Sea, a finer grid with a horizontal resolution of about 900 m (50” in longitude and 30” in latitude) and 25 vertical layers is nested into the coarse grid using a 2-way nesting. In the North Sea, the model configuration has a northern open boundary in the coarse mesh, which is closed with a sponge layer. Within this layer, the temperature and salinity are restored towards monthly mean climatological values (Janssen et al. 1999). A similar sponge region is included at the entrance to the English Channel. A two- dimensional model for the North East Atlantic, which is run separately by the BSH, provides information on external surges at the open boundaries. Tidal forcing is implemented using 14 tidal constituents and flooding and drying of tidal flats is applied (Bruening et al. 2014). The atmospheric forcing at the surface is based on meteorological forecast data provided by the German Weather Service (DWD). River runoff is prescribed as freshwater fluxes at the Fig. 1 Sea surface temperature on the 1st of April 2012 on the coarse (left) and fine (right) model domains. The coarse model grid excludes the region of the fine grid. In the left plot some geographic regions dis- cussed in the text are marked. Further, the yellow markers at 19.79? E, 62.725? N in the Gulf of Bothnia and at 27.54? E, 60.33? N in the Gulf of Finland show the location of profiles that will be discussed in Section 7