Combined model state and parameter estimation with an ensemble Kalman filter in a North Sea station 1-D numerical model
Ponsar, S.; Luyten, P.; Ozer, J. (2011). Combined model state and parameter estimation with an ensemble Kalman filter in a North Sea station 1-D numerical model. Ocean Dynamics 61(11): 1869-1886. dx.doi.org/10.1007/s10236-011-0477-5
In: Ocean Dynamics. Springer-Verlag: Berlin; Heidelberg; New York. ISSN 1616-7341; e-ISSN 1616-7228, more
|  |
| Keywords |
Data
Data assimilation
Estimation
Filters > Kalman filters
Models
Numerical models
Parameters
Profiles > Vertical profiles > Temperature profiles
Simulation
ANE, North Sea [Marine Regions]
Marine/Coastal |
| Author keywords |
Data assimilation; Ensemble Kalman filter; Combined model state andparameter estimation; Numerical modeling; North Sea |
| Abstract |
The combined estimation of model state and parameters is investigated by using an ensemble Kalman filter in a 1-D numerical study of the temperature profile at a North Sea station located at 55°30’ North and 0°55’ East. This simplified model implementation allows to test many configurations for the sampling of the model error for the model state as well as for the parameters. Temperature profiles from thermistor data are assimilated. The influence of a vertically or non-vertically correlated model error is examined through the computation of correlation coefficients, root mean square errors, and model bias. As the CS station is located in the North Sea region stratified in summer, vertically non-correlated model error terms have a slight positive impact on the assimilative runs. The benefit of the combined estimation of the model state and parameters is examined by comparison of a simulation where the model parameters are not adjusted to simulations with the combined estimation of the model state either with one parameter or with a set of parameters related to the surface heat exchange. The optical attenuation coefficient, the sensible, and latent heat exchange coefficients are considered. The best results are obtained when a set of parameters are simultaneously adjusted. |
|
