|
As a number of operational centers have implemented
four-dimensional Ensemble-Variational (4D EnVar) Data Assimilation (DA) for their global models,
the Developmental Testbed Center (DTC) continues to
conduct testing and evaluation of the GSI (Gridpoint
Statistical Interpolation) 4D hybrid EnVar system
for regional 3-km High Resolution Rapid Refresh (HRRR), as part of the
efforts to improve the convective scale and cloud resolving numerical weather
predictions at the National Oceanic and Atmospheric Administration Earth
System Research Laboratory (NOAA/ESRL).
Due to computational constraint, the operational HRRR 3km domain has been
reduced from the CONUS (Continental US) to the central US. The period for the
testing and evaluation is set to be September 3-9 of 2016, with hourly update
of the initial and boundary conditions from the retrospective 13-km Rapid
Refresh (RAP) runs. In addition to the experiments with the hybrid GSI
three-dimensional EnVar (3D EnVar)
as in the operational HRRR configurations, the hybrid 4D EnVar
is applied to the reduced HRRR domain to investigate whether the 4D hybrid EnVar system improves upon the performance of the
benchmark HRRR. Two-hours pre-forecast runs from each cycle are conducted to
provide GSI backgrounds at three different time levels (t-1hr, t and t+1hr)
to match the observations in three time bins. The 80-member global ensemble
forecasts at three time levels provide time-variant, flow dependent
background errors to the GSI in addition to the climatology background
errors. Preliminary results suggest that the hybrid 4D EnVar
GSI analysis gives better match between the observations and the HRRR
background than the 3D EnVar in wind and humidity.
Additional experiments with regional 3-km ensemble forecasts in the hybrid 3D
EnVar as compared to the 3D EnVar
using the low-resolution GFS ensemble suggests the potential benefit of
applying high resolution regional ensembles in the 4D EnVar
for better background error representative and therefore better forecasts.
|