Ensemble based data assimilation of bio-geochemical profile data in the Baltic Sea with a Nemo-ERGOM model system

The presentation focuses on two key points in bio-geochemical ensemble based data assimilation: the ensemble generation methods as well as the impact of the data assimilation on various bio-geochemical processes. We conduct the data assimilation experiments from October 2014. These are preparations for a future reanalysis product provided by the Baltic Monitoring and Forecasting Centre (BAL-MFC), which will cover the years 1993 – 2021. The Local Error Subspace Kalman Transform Filter (LESKTF) algorithm in the Parallel Data Assimilation Framework PDAF (http://pdaf.awi.de) is applied for data assimilation of profile data from the SHARK database on a daily basis. After the daily analysis is performed, the mean of the ensemble members is used in the Nemo-ERGOM model system. This method is used operationally by the BAL-MFC.

Effects of the number of ensembles, transformations, generation techniques and deflation of the ensemble are explored and verified in our ensemble generation studies. Rank histograms, skewness and kurtosis of the ensembles before and after assimilation are computed.

The influences of dissolved oxygen profile data assimilation on the nutrients in deep layers are studied and compared with the integral influences of univariate data assimilation of dissolved oxygen, nitrate, phosphate and ammonium on the same variables. Validation results of the univariate data assimilation scheme are presented and discussed in regards to quality enhancement for the future reanalysis product.