LAUSR

Abstract We develop an efficient data assimilation system that aims at quantifying the uncertainties of various biogeochemical states and parameters. We explore the use of four different ensemble estimation techniques for tuning poorly constrained ecosystem parameters using a one-dimensional configuration of the Ocean Biogeochemical General Circulation Model. The schemes are all EnKF-based operating sequentially in time but have different correction equations. The 1D model is used to simulate the biogeochemical cycle at three different stations in mid and high latitudes. We assimilate monthly climatological profiles of nitrate, silicate, phosphate and oxygen in addition to seasonal surface pCO 2 data, between 2006 and 2010. We use the data to optimize eleven ecosystem parameters in addition to all state variables of the model, describing the dynamical processes of the water column. Our assimilation results suggest the following: (1) Among all tested schemes, the one-step-ahead smoothing-based ensemble Kalman filter (OSA-EnKF) is robust and the most accurate, providing consistent and reliable state-parameter ensemble realizations. (2) Given the large uncertainties associated with the ecosystem parameters, estimating only the state variables is generally inconclusive and biased. (3) The OSA-EnKF successfully recovers the observed seasonal variability of the ecosystem dynamics at all stations and helps optimizing the parameters, eventually reducing the prediction errors of the nutrients’ concentrations. (4) The estimates of the parameters may have some temporally correlated features and they can also vary spatially between different regions depending on the magnitude of the bias in the observed variables and other factors such as the intensity of the bloom period. We further show that the presented assimilation system has the potential to be used in global models.