3C.2 Performance of the HWRF Analog Ensemble during the HFIP 2017 Real-Time Demonstration

Monday, 16 April 2018: 1:45 PM
Champions ABC (Sawgrass Marriott)
William E. Lewis, Univ. of Wisconsin–Madison, Madison, WI; and C. M. Rozoff, L. Delle Monache, and S. Alessandrini

The Analog Ensemble (AnEn) technique is a simple, inexpensive post-processing method for exploiting the relationships between model forecasts and observed quantities over long-periods of time. Given a set of historical simulations and a corresponding set of observations, the AnEn ingests a raw model forecast and produces a new forecast that accounts for systematic model behavior with respect to any observable quantity. Aside from its superior performance in first-moment estimates, the AnEn also provides a ready means of producing well-calibrated ensembles (i.e. second-moment estimates) and at a much lower computational cost than traditional dynamic model ensembles. It has been extensively tested for the probabilistic prediction of common meteorological variables (e.g., wind and temperature), air quality (ground-level ozone and particulate matter), for renewable energy applications (both solar and wind) and tropical cyclone (TC) intensity and intensity change (Vmax, DeltaVmax).

Using reforecasts of the 2014-2016 Atlantic and Eastern Pacific seasons conducted with the 2017 version of NCEP’s HWRF model, we have developed a new intensity-only AnEn model that leverages the predictions of the top-flight intensity models (i.e. the intensity consensus model IVCN). We will provide a summary of the intensity-only AnEn’s performance during the HFIP 2017 Demonstration (including hurricanes Harvey, Irma and Maria) and preview further extensions of the AnEn methodology to other aspects of the TC forecast problem.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner