Title: A dilemma of large-scales in LAM analysis

Author: Antonín Bučánek (Czech Hydrometeorological Institute)

We cannot sample background errors for scales larger than computational domains of limited area models. Elegant way how to manage this drawback is the use of the so-called digital filter blending which was developed in CHMI. It is a technique allowing to preserve large scale circulation structures from the host model analysis and small scale features resolved by the high resolution guess of the ALADIN model. The use of the digital filter blending in LAM is still beneficial even after introduction of data assimilation. We combine 3DVAR assimilation scheme with the digital filter blending for its relative simplicity and low computational cost. We learn the behaviour of the data assimilation system in various setups: digital filter blending, 3DVAR and proposed Blending-3DVAR setup. We evaluate forecast performance, especially quantitative precipitation forecast, for the above mentioned setups and for the resolutions of 4.7 km and 2.2 km (convection permitting scales). The heavy precipitation episodes affecting Central Europe in 2013 with their low predictability are well suited for comparison of the setups. For demonstration we namely use the extreme precipitation event of June 1 - June 3 2013 showing high sensitivity to initial conditions.

GMAO Head: Steven Pawson Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: Feb 9 2015