S2.8 How Reproducible are the ECMWF and GFS Hurricane Sandy Forecast Tracks?

Tuesday, 1 April 2014: 9:15 PM
Pacific Ballroom (Town and Country Resort )
Nick P. Bassill, University of Utah, Salt Lake City, UT; and E. Zipser

Hurricane Sandy (2012) followed a fairly unusual track, which included a turn toward the northwest while east of North Carolina, which allowed for a New Jersey landfall late on 29 October 2012. However, track forecasts made by operational models significantly earlier in Sandy's life cycle exhibited a notable bifurcation in forecast track. On 23 and 24 October, while Sandy was still located in the central Caribbean, the ECMWF model consistently forecast a northeastern United States landfall, while the GFS model consistently forecast Sandy to head out to sea. This was generally true of their respective ensemble members as well. Some suggested reasons offered in real-time for this bifurcation were that the models differed due to their respective differences in resolution or data assimilation schemes, among many other possibilities.

This study asks the question: Are the respective ECMWF and GFS track forecasts reproducible using a single dynamical core (WRF-ARW) and a single set of initial conditions? A series of experiments were conducted for the 23rd and 24th at both 0000 UTC and 1200 UTC in which every model parameter was identical, except for choice of cumulus parameterization. A pair of 7.5 day forecasts were created where one used the standard GFS cumulus parameterization (Simplified Arakawa-Schubert) while the second used one representative of the ECMWF cumulus parameterization (Tiedke). Despite the fact every forecast used the same initial and boundary conditions (based upon the appropriate GFS forecast), the forecasts all exhibited the same forecast track bifurcation as seem in operational models. This was also true across a wide variety of grid spacings. Collectively, this indicates that the cause of Sandy's track split was not due to differences in resolution or data assimilation schemes employed by the operation EVMWF and GFS models, but rather by model physics. The likely cause of these track differences will be further explored in this presentation.

- Indicates paper has been withdrawn from meeting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