Several methods for data assimilation and ensemble forecasting were compared by the use of quasi-geostrophic 3D-Var data assimilation simulation system. It was necessary to develop the adjoint of this model to compare 2D-Var with 4D-Var and with the local ensemble Kalman filtering (LEKF). The bred vectors (BV) were obtained through a breeding cycle which started by adding random perturbation to the analysis and the singular vectors (SV) were obtained by assuming that the perturbation behaved linearly with the chosen optimization time interval. The results show that the final SVs have a similar structure as the BVs and both of them have a ability to depict the shape of the background error.
|Number of pages||10|
|Journal||Bulletin of the American Meteorological Society|
|State||Published - 2004|
|Event||Combined Preprints: 84th American Meteorological Society (AMS) Annual Meeting - Seattle, WA., United States|
Duration: 11 Jan 2004 → 15 Jan 2004