When dealing with inverse modelling (Data Assimilation, DA) using variational techniques or statistical Monte Carlo based methods (at least for oceanographic applications) we pretty often know little about model uncertainties and data error statistics. The inverse solution, however, crucially depends on our prior assumptions on the error statistics. Kivman et al. (2001) suggested the entropy approach (PME) to tune the weights of model and data costs in the generalized inversion. Implementation of such an approach for a problem of state and parameter estimation in biogeochemical modelling allowed us to obtain reliable estimates of physiological parameters and, moreover, to inference about data quality and model uncertainties (Losa et al., 2004). Here we investigate a possibility of the PME application for calibrating an ensemble based DA system for an operational circulation model of the North and Baltic Seas.