Satellite radiance data assimilation with a limited-area ensemble Kalman filter and 3D-Var analysis system

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 26 January 2011: 11:30 AM
Satellite radiance data assimilation with a limited-area ensemble Kalman filter and 3D-Var analysis system
2B (Washington State Convention Center)
Craig S. Schwartz, NCAR, Boulder, CO; and Z. Liu, Y. Chen, and X. Y. Huang
Manuscript (349.5 kB)

Data assimilation (DA) techniques in limited-area deterministic and ensemble numerical weather prediction (NWP) model frameworks were employed to study Typhoon Morakot, which developed in the Western Pacific Ocean in August 2009 and made landfalls in both Taiwan and China. Over the lifetime of Morakot (~ 1 week), the Weather Research and Forecasting (WRF) model's three-dimensional variational (3DVAR) DA system generated analyses for the deterministic experiments while an ensemble Kalman filter (EnKF) produced analyses from a 64-member ensemble. The analyses initialized triple-nested 72-hr Advanced Research WRF (ARW) model forecasts (from the mean analysis in the ensemble experiments).

A particularly unique aspect of this study was the direct assimilation of microwave satellite radiances with 3DVAR DA and the EnKF. Although radiance assimilation is an important component of many operational global analysis systems, the impact of assimilating radiances within a regional domain is less studied and understood. Thus, output from our configurations with radiance DA was compared to output from parallel experiments where radiances were withheld from the observational datastream. This design permitted a robust assessment of limited-area NWP analysis and forecast sensitivity to radiance assimilation for a tropical cyclone case.

Results suggested that assimilating radiances within both the deterministic and ensemble settings yielded better intensity forecasts of Morakot compared to the experiments where radiances were not assimilated. However, the impact of radiance DA on track forecasts was more ambiguous. Precipitation forecasts from the experiments and pros and cons of using an ensemble versus deterministic forecasting system for this case study will also be discussed.