A preliminary comparison of GEOS-GSI with GEOS-JEDI using the complete GEOS observing system.

Transitioning the NASA GMAO GEOS data assimilation capabilities to JEDI involves replacing the GEOS Gridpoint Statistical Interpolation (GSI) with a JEDI-based analysis. Performing fair comparisons between the corresponding GEOS-JEDI system and the present GEOS-GSI has involved downgrading the latter's capabilities to compensate for missing capabilities in the former. As the capabilities in JEDI mature and start matching those of GSI's experimentation has evolved from simple 3DVAR using only radiosondes, to adding satellite observations without VarBC, to now having GEOS-JEDI use hybrid 4DEnVar, with full observing system, VarBC and most of the knobs typically used in the present operational (and experimental) version of GEOS-GSI. This presentation will provide a discussion of the process and path taken to replace GSI with JEDI in GEOS. The presentation will show latest results, and more importantly, it will tackle what is being done to address the missing, and necessary, knobs to allow GMAO to fully transition to a JEDI-based atmospheric data assimilation system, including the addition of a Tangent Linear Normal Mode Constraint (TLNMC) that works in conjunction with the GSI background error formulation used in GSI and made compatible to run within the JEDI framework.