GSI-based Hybrid Ensemble-Variational Data Assimilation for Global, Hurricane and Storm-scale Numerical Weather Prediction
Xuguang Wang
University of Oklahoma
8 August, 1 pm in 2155
Abstract:
GSI-based hybrid ensemble-variational data assimilation (DA) system
including both 3DEnVar and 4DEnVar improved US NWS global forecast
significantly. This seminar will discuss our recent research and
development of the system for a variety of
scales including global, hurricane and storm scales.
A method which effectively increases the size of the ensemble incurring
minimum cost is implemented in GSI-based 4DEnVar. Experiments with GFS
configured similarly to the 4DEnVar pre-implementation test were
conducted. It was found that the GFS forecast was further improved.
Research and development have also been made to extend the GSI-based
hybrid DA system for regional modeling systems. The hybrid EnKF-Var DA
system is extended with the operational HWRF modeling system to improve
high-resolution, convection-allowing tropical cyclone prediction. An
end-to-end, continuously cycled hybrid DA system was developed using a
newly developed directed moving nest strategy. Experiments were
conducted for hurricane Edouard (2014) in which all operational
observations including conventional in-situ data, satellite wind,
tcvital, satellite radiances, and tail Doppler radar observations were
assimilated. It was found that: a) the dual resolution hybrid DA
improved upon the coarser, single resolution hybrid; b) vortex
initialization and relocation in the control and relocation of the
ensemble background improved the forecasts; c) using 4DEnVar DA in the
TDR-involved cycles improved the intensity forecasts for early lead
times compared to 3DEnVar DA; and d) the hybrid system improved
intensity forecasts relative to operational HWRF during the
intensification period due to the alleviation of the “spin-down” issue
because the hybrid better analyzed the structures of an intensifying
storm. Issues associated with the vortex initialization will also be
discussed.
The GSI-based hybrid DA system is also extended with the convection
resolving WRF ARW and NMMB models for multi-scale assimilation of both
conventional in-situ observations and radar observations to improve
convective scale hazardous weather forecasts over the Continental US
(CONUS). Systematic comparison of GSI based 3DVar and EnKF was first
conducted in the context of multiscale data assimilation where scales
ranging from convective scales to synoptic scales were resolved by both
the model and the observations. The cases include many examples of both
discrete cellular convection and organized Mesoscale Convective Systems
(MCSs). It was found that convective scale precipitation forecasts
initialized by GSI-based EnKF were much more skillful than GSI-based
3DVar. The positive impact of assimilating radar data lasted much
longer in EnKF than 3DVar. Design, implementation and performance of
the GSIbased convection-allowing ensemble based data assimilation and
forecast system in support of the 2015 PECAN field experiment will also
be discussed.
In addition, a new method to directly assimilate reflectivity
observations was proposed and implemented in the GSI-based EnVar
system. This new method avoids the use of the tangent linear and
adjoint of the nonlinear reflectivity operator and therefore overcomes
the issues of using hydrometeors or logarithm of hydrometeors in GSI
variational minimization. The newly proposed method is examined on the
analysis and prediction of the 8 May 2003, Oklahoma City, tornadic
supercell storm at 2km model grid. Both the probabilistic forecast of a
strong low-level vorticity and the maintenance of the strong updraft
and vorticity in the new method are more consistent with the reality
than in option-logarithm and option-mixing-ratio. Detailed diagnostics
suggest that a more realistic cold pool due to the better analyzed
hydrometeor mixing ratios in the new method than in other methods leads
to the constructive interaction between the surface gust front and the
updraft aloft associated with the mid-level mesocyclone.
Challenges, implications and future research under the NGGPS paradigm will be discussed at the end of the seminar.