Operational Applications of Ensemble Sensitivity Analysis: What is the Origin of the Forecast Uncertainty during Winter High-impact Weather Events?

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 4 February 2014
Hall C3 (The Georgia World Congress Center )
Minghua Zheng, SUNY, Stony Brook, NY; and E. K. M. Chang and B. A. Colle

Ensemble forecasts are appealing for operational forecasters because they can provide the ensemble spread or probability of a certain kind of weather at a particular time and for a particular location. However, ensemble outputs are often not used effectively mainly because of the lack of a simple and quantitative tool to assess the forecast uncertainty and its origin as well as an efficient way to verify the real-time ensemble features. Previous studies have shown that ensemble sensitivity analysis, which employs a linear correlation and regression between a chosen forecast metric and the forecast state vector, can be used to analyze the forecast uncertainty development for both short- and medium-range forecasting. This presentation further applies the ensemble sensitivity tool to real-time ensemble outputs by summarizing three ways (described below) that this method can be applied through three case studies of winter high-impact weather events.

To elucidate the application of the ensemble sensitivity tool, a US East Coast snowstorm on 26-28th December 2010 is first examined using the 50-member European Center for Medium-Range Weather Forecasts (ECMWF) ensemble data. Sensitivity based the Principal Components of the leading Empirical Orthogonal Functions (EOF) patterns on the 5.5-day ensemble forecast, as well as the ensemble mean sea-level pressure difference between two run cycles (forecast jump between the 3.5- and 2.5-day forecasts) are calculated. These identified the sensitive regions and weather systems at earlier times associated with the cyclone intensity and track uncertainty separately. Specifically, both approaches suggest that the cyclone track's shift is linked with the uncertainties in the short wave trough over the southern Great Plains. The sensitivity approach is then run forward in time using forward ensemble regression based on short-range forecast errors, which further confirms that the short-term error over the Southern Plains trough was associated with the shift of cyclone position between the two forecast cycles. In addition, two recent winter high-impact events: a western US heavy rain and snow event on November 29th December 5th 2012 and a central and eastern US winter storm on March 25th 2013 have also been analyzed using the sensitivity tool based on The Observing System Research and Predictability Experiment (THORPEX) Interactive Global Grand Ensemble (TIGGE) multi-model EPSs data. Using these three cases, the accuracy of sensitive regions identified by ensemble sensitivity tool is investigated using the leave-one-out cross validation (LOOCV) method.