A Hybrid Data Assimilation System (WRF-VAR and Ensemble Transform Kalman Filter) Based Retrospective Tests with Real Observations

Ensemble-based analysis data assimilation techniques provide

a good avenue for addressing uncertainties at the initial conditions.

In this study, we have made real data based extensive retrospective

runs with Weather Research and Forecasting Variational Data Assimilation

System-Ensemble Transform Kalman Filter hybrid DA system (hereafter

WRFVAR-ETKF hybrid DA system) that gives flow-dependent estimate

of background forecast.

We have implemented WRFVAR-ETKF hybrid DA system at National Center

for Atmospheric Research, Data Assimilation Testbed Center -where

various data assimilation techniques and new observation types

are tested and evaluated for research and operational usage. In our

WRFVAR-ETKF hybrid DA system, the ETKF technique has been used to

update ensemble perturbations and hybrid technique is for updating

ensemble mean. In a cycling mode; both initial and boundary conditions

are updated and used as input to run WRF for generating next cycle's

ensemble members.

We have designed our run domain to include also the Caribbean area

(5-35N, 100-50W). Our Caribbean domain has 45-km horizontal resolution

and 57 levels in the vertical. Some extensive retrospective runs with

3-hourly cycling have been conducted for 30-day period (20070815-20070915)

- which also covers the time of the tropical cyclone Dean (TC Dean) that

developed into a category five hurricane. Conventional observations

and Global Forecasting System model data sets used as initial input data.

The presence of TC Dean in our run domain-period also presented

an opportunity to examine how non-localized ETKF technique could

perform. We have tested two versions of ETKF system: the first

version produced modest inflation factors and provided stable runs,

but less ensemble spread; while the second version gave higher inflation

factors and more ensemble spread, but a few CFL issues in some WRF runs.

Both tests have provided encouraging results in terms of the impact

of flow-dependent capacity of WRFVAR-ETKF hybrid DA system and

improvement over the test run that only used WRF-VAR.