HRD seminar – Dr. Jason Sippel, NCEP/EMC, College Park, MD – 30 March 2016

Dr. Sippel presented a seminar on “The Challenges in Developing the Operational HWRF DA System:  What I’ve learned after 1.5 years at EMC”.


Though the HWRF model has seen tremendous advances over the past decade, largely due to HFIP, the HWRF data assimilation system still lacks what is needed to properly analyze tropical cyclone vortices.  The methods necessary to initialize hurricanes were developed nearly a decade ago, but they still have not been operationally implemented.  The most severe deficiency of the current system is reliance upon background error covariance either from the NCEP global ensemble, or from a 6-h HWRF forecast initialized with the global ensemble.  Biases in model physics also appear to interact with HWRF initialization procedures and cause severe problems with HWRF data assimilation.  The result of these issues is that HWRF suffers extreme spindown when data assimilation is performed in the inner core of many tropical cyclones, which requires that the model not use most reconnaissance data within the storm.  Instead, a vortex initialization procedure, which has its own deficiencies that will be discussed, is performed for the inner core region. 

Several developments are necessary to give HWRF a truly effective data assimilation system for all tropical cyclones.  First, error covariance needs to be reflective of mesoscale uncertainty, which can only be achieved by cycling of HWRF ensemble perturbations.  Additional developments, including high-frequency cycling, and both pure ensemble and 4D-hybrid data assimilation need to be explored within the context of the operational system.  Furthermore, it is necessary to eliminate model biases that are producing inconsistencies between the model first guess and observations.  After these problems have been addressed, there is also a need to explore ways to circumvent the current vortex initialization procedure.  Some of these developments are ongoing, and others will require significant investment from HFIP partners and perhaps a re-examination of the way HFIP has approached data assimilation. 

A recording of the presentation is available on the anonymous ftp site:

Screen Shot 2016-03-30 at 12.46.32 PM