Recently, the numerical biogeochemical models for the Baltic sea are well developed in improving resolution and the simulation accuracy. Furthermore, the coverage and amount of the biogeochemical observations have become large enough, especially in the Baltic subsurface. Therefore, the confidence for evaluating the Baltic biogeochemical state becomes high. It is of interest and importance to estimate the variability of biology variables and its associated uncertainty over a long time period and estimates the limits of applicability of long-term reanalyses or reconstructed data.

A multi-year physical-biogeochemical state is reconstructed in the Baltic Sea during the period 1993-2016. The model system is based on the Swedish coastal and ocean biogeochemical model (SCOBI) coupled to the NEMO-Nordic circulation model. A weak coupled data assimilation system based on a sequential ensemble interpolation kalman filter (SEIK) is adopted to merge the information of model and observations. The high resolution remote sensing sea surface temperature, salinity and temperature profiles, oxygen and nutrient profiles measurements are assimilated into the NEMO-SCOBI. In order to produce consistent analysis, both physical and biogeochemical observations are assimilated in the same time. Based on the numerical experiment with and without data assimilation, the inter-comparison is implemented to validate the reanalysis results. The model can reproduce the variability of ocean and biogeochemical parameters in important dynamic processes (e.g. the inflow process). However, the model simulation shows obvious biases, especially in the deep layers. As expected, the assimilation process is found to impart significantly positive impacts on the physical and biogeochemical simulation. For example, with the comparison with observations, the oxygen, nutrients, temperature and salinity biases in reanalysis at the Baltic proper are significantly reduced relative to reference run. Furthermore, the reanalyzed hypoxia in the Baltic Sea is better than the free run relative to the observed one. The results denote that as a consistent reconstructing of physical and biogeochemical state for the Baltic Sea, the output data set of this reanalysis can inform the management of the Baltic Sea ecosystem. For example, the analysis or prediction of the trend and reason of oxygen deficiency in the Baltic water and its potentially threatening. Moreover, for projections of future climate and for nutrient load abatement scenario simulation, this reconstructed record has very high scientific value as a reference data set for the historical period of the climate simulations.