Impact of the vertical coordinate system on the Baltic Sea hydrography

The Baltic Sea is a semi-enclosed brackish marginal sea in Northern Europe. It is a sensitive system where increased anthropogenic pressure has led to a poor ecological status, with resulting hypoxia and increased algal blooms. The circulation, hydrography and stratification are strongly constrained by the bathymetric features of the region and the net freshwater input. The shallow entrance region limits the inflow of oceanic water from the North Sea, while the net freshwater input drives an estuarine-like circulation that leads to an outflow-dominated net exchange. Episodically inflows can occur through the connecting straits. When these inflows are substantial enough, during so-called Major Baltic Inflows (MBIs), they can ventilate the deeper parts of the Baltic Sea and oxygenate the deepwater. One of the most important design choices for an ocean model is the vertical coordinate system. Traditionally most regional Baltic Sea models have employed z-coordinates, although both terrain-following and adaptive vertical coordinate models also exist. Here we use a NEMO-based configuration (NEMO-Nordic) to study how the choice of the vertical coordinate system impacts the Baltic Sea hydrography and the representation of MBIs. A recently developed Multi Envelope s-coordinate (MEs) configuration is compared to two versions of a well-established z-coordinates with partial steps (zps and zps_bfr) configuration. MEs and zps use a constant roughness length scale (z0), while zps_bfr has a tuned horizontally variable z0. Overall, the MEs configuration shows very promising results compared to the zps/zps_bfr configurations. Based on observations from a number of monitoring stations we find generally lower error metrics for MEs, particularly for salinity. Both the MEs and zps_bfr configurations represent the stratification and MBIs well, while the zps configuration has a large positive bias in salinity, indicating too strong inflows. Our results show that one can achieve similar results with MEs and zps-type configurations, provided that the model using geopotential levels employs a significantly ad-hoc tuned bottom friction. In the latter case, however, such a tuning has no physical justification and involves a quite time-consuming series of sensitivity experiments and computational resources. Furthermore, we find that MEs has a more dynamical and well-resolved bottom boundary layer which at some locations leads to higher variability in near-bottom salinities and temperatures compared to zps/zps_bfr and observations.