The ensemble sensitivity method of Liu and Kalnay (2008) is applied to the WRF-LETKF system, and impacts of real observations on short-range forecasts are assessed with additional enhancements of considering localization and introducing a targeted area that are essential for the high-dimensional real application. This ensemble-based method achieves the same goals as the adjoint-based method (Langland and Baker 2004) but without using an adjoint model.

The results in the case of Typhoon Sinlaku (2008) show that upper-air soundings have the largest impact on improving 12-h forecasts and that the targeted impact evaluation performs as expected with shorter computational time. Denying negative-impact observations actually improves the forecasts, validating the estimated observation impact.