12.4 Impact of TAMDAR Data on Hurricane Paula with an Eigen-Structure Dependent Inflation Scheme in the WRFDA/ETKF Hybrid Data Assimilation System

Thursday, 10 January 2013: 2:15 PM
Room 9C (Austin Convention Center)
Dongmei Xu, UCAR, Boulder, CO; and A. P. Mizzi and X. Y. huang
Manuscript (557.6 kB)

This paper documents application of the NCAR/MMM WRFDA/ETKF hybrid data assimilation system with six-hour cycling to hurricane forecasting on the AirDat, LLC operational grid with TAMDAR observations and an improved (Wang et al., 2007) inflation scheme.

Earlier work showed that that the inflation scheme was not well behaved when there were large variations in the number of ETKF observations from one cycle to the next. That problem becomes acute with six-hour cycling.

The ETKF transform vector depends on the eigen-structure of the observational space sum of the squared ensemble perturbations of the expected observations normalized by the observation errors. The magnitude of the ETKF spread contraction in any particular eigen-direction is inversely related to the associated eigenvalue. The improved scheme makes the inflation factor eigenvector-dependent by basing the -factor on the projection of the sum of the squared innovations onto the particular eigenvector. As a result, it restores greater amounts of covariance in the eigen-directions of the greatest covariance reduction. The preliminary results show that the revised inflation factor scheme ameliorates the ETKF observation number instability by maintaining greater inter-cycle continuity for the eigen-structure of the posterior spread.

TAMDAR is a mulit-function in-situ atmospheric sensor for aircraft. TAMDAR sensors are installed on commercial aircraft and continuously transmit atmospheric observations. Previous studies have shown that including TAMDAR observations during data assimilation improves forecast skill. This is one of the first studies to apply TAMDAR data to hurricane forecasting. The results show improved six and twelve-hour forecast skill compared to not using TAMDAR data. The results also show that TAMDAR observations improve the forecast skill for hurricane parameters such as track location and error, central pressure, and maximum wind.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner