Assimilation of surface winds from satellites: present and future

In this presentation, we investigate the ability of an operational data assimilation scheme, the NCEP Gridpoint Statistical Interpolation (GSI) to handle wind vectors derived from the QuikSCAT satellite, and propose future strategies for improvement.

In many situations, the influence of assimilating QuikSCAT wind vectors on NCEP Global Forecast System (GFS) tropical cyclone analyses and forecasts is minimal. This is perhaps due in part to the strict quality control, and the 'super obbing' algorithm that removes smaller-scale structures that may be important near tropical cyclones. At least as important is the background error covariance structure in the data assimilation. With a quasi-isotropic, flow-independent error covariance specification, the correction to the tropical cyclone analysis is found to be mostly geostrophic, and significant only in the lower troposphere. Modifications to the tropical cyclone and environment at higher levels are minimal. The influence of altering the vertical background error covariance in the GSI on analyses and forecasts of tropical cyclones will be presented.

The introduction of anisotropic and flow-dependent error covariance information reveals a more significant influence of the surface wind observations throughout the vertical extent of the tropical cyclone, with a horizontal correction that departs from geostrophy. The potential for improved analyses using an Ensemble Kalman Filter in global and regional models will be illustrated.