Hybrid 3dVar assimilation system in its quasi-operational suite. This system supports various NASA
missions as well as it serves as the sanity check for updates that eventually make up the bulk of
a system to be promoted to supporting enhanced reanalysis efforts at GMAO. An upgraded parallel version of
this system has been running since mid-2016 that has as its main highlights an increased horizontal resolution
and an upgrade from hybrid 3dVar to a hybrid 4d ensemble-variational (4dEnVar) procedure. The GMAO
hybrid assimilation systems entail the following components: (i) a central, high-resolution,
forecast (background) using the GEOS-5 (FV3-based) general circulation model; (ii) a
hybrid 4dEnVar analysis, using the Gridpoint Statitical Interpolation (GSI) system; (iii) an ensemble of
low resolution GEOS-5 model background integrations; and (iv) an ensemble analysis produced using
a square-root ensemble Kalman filter (EnKF) procedure. The upcoming system, running in the parallel
suite, increases the effective resolution of its central forecasts from 28 km to roughly 12.5 km; its
hybrid 4d GSI analysis is performed at 28 km (in constrast to the quasi-operational 50 km hybrid 3dVar analysis);
and it brings in an increased ensemble horizontal resolution from roughly 100 km in the present
quasi-operational system to 50 km.
The increased model and analysis resolutions combined with the advanced assimilation strategy results in
non-negligeable improvements in different aspects of the system. As expected, the most direct benefit from
increased horizontal resolution is seen when analyzing and predicting tropical storms, with improvements in both
track and intensity forecasts. Less obvious is the considerable reduction in tropical forecast errors
brought about by the upgraded system, with tests indicating those to be due to the four-dimensional aspect of
the analysis.
Our presentation will provide multiple illustrations of improvements in the upcoming GMAO four-dimensional
system, as well as provide a view for further develop.