The impact of including a weak dynamical constraint in the GEOS-5 Data Assimilation System

A weak dynamical constraint is applied to the GEOS-5 Data Assimilation System of the Global Modeling and Assimilation Office at GSFC/NASA aimed at reducing the gravitational noise in the analysis products. The GEOS-5 data assimilation system consist of two main components: the Goddard GEOS-5 general circulation model and the Grid-point Statistical Interpolation System (GSI) developed at NCEP. The weak constraint is included by adding another term, called Jc, to the cost function to be minimized by the analysis scheme. This Jc is a quadratic measure of weighted squared differences between analyzed and background tendencies of both horizontal wind divergence and ageostrophic vorticity. It uses a vertical mode decomposition and selection together with a horizontal scale-dependent rescaling of the tendencies, analogous to those used in traditional normal mode initialization. Both of these modifications are necessary to obtain adequate convergence of the minimization algorithm and appropriate noise reduction. Two sets of cycled analyses and forecasts are performed with and without the Jc term. We will present diagnostics of the tendencies both in spatial and spectral spaces. We also present a comprehensive examination and assessment of the experimental results, including comparisons of forecast skill, initial tendency noise reductions in subsequent forecasts, statistics of fits of both analysis and background fields to observations, and monthly mean fields. All comparisons show that including the constraint significantly improves the analysis. The relationship between noise as measured by the JC term and other, including normal mode based, measurements of gravitational noise will be discussed.