Development and evaluation of a multi-functional data assimilation testbed based on EnKF and WRFVar with various hybrid options

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015: 2:30 PM
131AB (Phoenix Convention Center - West and North Buildings)
Yonghui Weng, Pennsylvania State University, University Park, PA; and F. Zhang, J. Poterjoy, Y. Ying, C. Melhauser, D. Tao, S. J. Greybush, X. Zhang, and J. Sun

The PSU WRF-based EnKF hurricane analysis and prediction system had been operated successfully in real-time since 2008. Based on this cycling EnKF and the NCAR WRFDA, the newly established Penn State Center for Advanced Data Assimilation and Predictability Techniques (ADAPT) has been developing a unified multi-functional data assimilation testbed through integrating 3DVar, 4DVar, EnKF, and various hybrid approaches (E3DVar, E4DVar and 4DEnVar). The unified system can be used for both research and real-time operational purposes, and for intercomparison of various hybrid and coupling data assimilation approaches with different forecast models (e.g., WRF-ARW, HWRF and COAMPS). In this study, each of the 6 advanced data assimilation methods implemented so far in the testbed will be evaluated by using the approach of continual cycling hurricane analysis and prediction introduced in Weng and Zhang (2014). The inter-comparison among different schemes and models is used to: 1) evaluate the impacts of hurricane inner-core observation assimilation with different data assimilation technique; 2) to investigate the role of the stationary covariance in hybrid data assimilation methods in the presence of pre-existing ensemble covariance tuning parameters; 3) to compare hybrid methods systematically over a variety of data assimilation scenarios; 4) to investigate the role of adjoint model in the advanced data assimilation technique by comparing the adjoint-based E4DVar versus the ensemble-based 4DenVar and 5) to distinguish the impact of initial condition versus model uncertainties in hurricane forecasts. After initial development and evaluation, we intend to make the testbed as an open-source platform available to the community for research, educational and/or operational uses, as well as for feedback and user contributions.