Error analysis of a highresolution physical model of the Mediterranean Sea
Vandenbulcke, L.; Barth, A.; Alvera Azcarate, A.; Lenartz, F.; Rixen, M.; Beckers, J.M. (2007). Error analysis of a highresolution physical model of the Mediterranean Sea. Université de Liège: Liège. 1 poster pp.
Is gerelateerd aan:Vandenbulcke, L.; Barth, A.; Alvera Azcarate, A.; Lenartz, F.; Rixen, M.; Beckers, J.M. (2007). Error analysis of a highresolution physical model of the Mediterranean Sea, in: Mees, J. et al. (Ed.) VLIZ Young Scientists' Day, Brugge, Belgium 2 March 2007: book of abstracts. VLIZ Special Publication, 39: pp. 66, meer

Trefwoorden 
Errors > Analytical errors Models > Scale models Mediterranean [Marine Regions] Marien 
Evenement  Top  Auteurs 
 7th VLIZ Young Marine Scientists' Day 2007

Auteurs   Top 
 Vandenbulcke, L.
 Barth, A.
 Alvera Azcarate, A.

 Lenartz, F.
 Rixen, M.
 Beckers, J.M.


Abstract 
We analyze the errors that are inevitably associated to hydrodynamic models, in a realistic case. The error of the GHER model in the Mediterranean Sea has already been studied in e.g. Beckers et al. (2000) by comparing it with other primitive equation models, or in Alvera (2004) by comparing the model with observations and with the climatology, using usual statistical methods and also wavelet decompositions. In this study, we rather study the sensitivity of the model to various variables using an ensemble of models. We chose a relatively high resolution, 1/16°, corresponding to the resolution now used in operational OGCMs covering the Mediterranean, such as the MFS system (http://www.bo.ingv.it/mfs). We explain how we generated an ensemble of model simulations, where various moreorless well known inputs are allowed to vary according to the uncertainty affecting them. Statistics calculated on this ensemble are, in fact, the response of the nonlinear hydrodynamic system to errors on the forcing terms. When those statistics are calculated at a certain timestep, they allow us to provide a spatial analysis of the model error; statistics calculated over the time dimension will show whether errors are intensified by the system, or rather disappear. The model error is interesting as such. However, it can also be used for different purposes. For example, it allows using data assimilation techniques without needing the usual assumptions of reducedrank Kalman Filters. It also allows studying the sensitivity of coupled models (biological, oil spill, searchandrescue, …) to physical forcings. 
