A sequential ensemble prediction system at convection-permitting scales
- Author(s)
- Marco Milan
- Abstract
A sequential data assimilation approach (SAM) that incorporates elements of particle filtering with resampling (SIR, Sequential Importance Resampling) is introduced. SAM is applied to the COSMO-DE-EPS, which is an ensemble prediction system for weather forecasting on convection-permitting scales. At the convective scale and beyond, the atmosphere increasingly exhibits non-linear state space evolutions. For an ensemble-based data assimilation system, this requires both an adequate metric that quantifies the distance between the observed atmospheric state and the states simulated by the ensemble members, and a methodology to counteract filter degeneracy, i.e. the collapse of the simulated state space. We, therefore, propose a combination of resampling, which accounts for simulated state space clustering, and nudging. SAM differs from the classical SIR approach mainly in the weighting applied to the ensemble members. By keeping cluster representatives during resampling, the method maintains the potential for non-linear system state development. With three convective case studies, we demonstrate that SAM improves forecast quality compared with the control EPS (EPS without data assimilation) for the first 5–6 h of forecast.
- Organisation(s)
- Department of Meteorology and Geophysics
- External organisation(s)
- Rheinische Friedrich-Wilhelms-Universität Bonn
- Journal
- Meteorology and Atmospheric Physics
- ISSN
- 0177-7971
- DOI
- https://doi.org/10.1007/s00703-013-0291-3
- Publication date
- 2013
- Peer reviewed
- Yes
- Austrian Fields of Science 2012
- 105206 Meteorology
- Portal url
- https://ucrisportal.univie.ac.at/en/publications/fc54fee0-66f9-4586-8ece-b63aec0ad66e