Hybrid-Mixed Lognormal-Gaussian 3D and 4D VAR
Abstract
A major assumption that is made in variational, ensemble, and hybrid data assimilation systems is that all of the errors that are minimized are Gaussian distributed random variables. Theory and applications have been developed at the Cooperative Institute for Research in the Atmosphere (CIRA) at Colorado State University (CSU) that allows for the minimization of a mixture of Gaussian and lognormal distributed errors simultaneously.
The theory mentioned above has been developed for the static versions of full field and incremental 3D and 4D VAR. In this presentation we shall introduce the theory for the hybrid versions for the full field versions of 3D and 4D VAR and compare the static to the hybrid versions with the Lorenz 1963 model to show the impact of the introduction of flow dependency into the background error covariance matrix which the mode of the mixed distribution is a function of.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMNG21B0921F
- Keywords:
-
- 3315 Data assimilation;
- ATMOSPHERIC PROCESSES;
- 0555 Neural networks;
- fuzzy logic;
- machine learning;
- COMPUTATIONAL GEOPHYSICS;
- 1640 Remote sensing;
- GLOBAL CHANGE;
- 3275 Uncertainty quantification;
- MATHEMATICAL GEOPHYSICS