Data Assimilation Seminar
Dr. Tobias Necker, (July 3, 2023, 15:30-17:00)
|Affiliation||University of Vienna, Austria / R-CCS|
|Title||The Potential of Large Ensemble Simulations: Insights from 5 Years of Research with a 1000-Member Ensemble|
This seminar will present highlights and outcomes from a collaborative research effort between RIKEN, the Ludwig-Maximilians-Universität in Munich (Germany), CIMA University of Buenos Aires (Argentina), and the University of Vienna (Austria). The collaboration aimed to improve and better understand the predictability and data assimilation of high-impact weather events by applying a convective-scale 1000-member ensemble simulation. The large ensemble simulation was conducted in 2018 on the K-computer and has been applied to study various aspects ranging from error covariances over observation impact to forecast verification.
The first part of the seminar will provide an overview of the conducted research and present selected highlights from 5 years of exciting research. The second part of the seminar will focus on the potential of large ensembles to better understand error covariances and sampling errors on convective scales, which are decisive for the success of ensemble and hybrid data assimilation systems. As demonstrated recently by Necker et al. 2023, the empirical optimal localization (EOL) method allows for deriving an optimal localization from a large ensemble dataset. In detail, we will discuss (1) Vertical error correlations and EOL estimates for different variables and settings. (2) The effect and behavior of the EOL compared to common localization approaches, such as distance-dependent localization with a Gaspari and Cohn function. (3) How to achieve positive definiteness of localization. (4) Optimal vertical observation space localization of cloudy infrared and visible satellite observations.