Improving the Ensemble Predictability of High-impact Events using Forecast Sensitivity

Ensemble prediction using an ensemble Kalman filter involves the propagation of a number of forecasts begun from a distribution of analyses that characterizes the uncertainty of the atmospheric state. These forecasts then provide a probability distribution of the future atmospheric state, and provide forecasters valuable information regarding the uncertainty of specific, high-impact weather events. By calculating the forecast sensitivity (ensemble or adjoint) of specific forecast events, it may be possible to identify an ensemble subset that improves the prediction of these events over the full ensemble itself. The hypothesis tested here is that by weighting forecast error in sensitive regions associated with specific forecast aspects, ensemble subsets can be found that improve forecasts of these aspects. This study uses an observing system simulation experiment (OSSE) in a perfect model environment to realize these benefits for land-falling, mid-latitude cyclones on the west coast of North America over a full winter season. Further research within a more realistic framework is discussed based on the results found here.