June 17, 2010 Meeting Summary
Dan Iredell presented an introduction to HYCOM in the HyHWRF (HYCOM/HWRF) system. Dan first gave an overview of the RTOFS domain and the HYCOM subdomains. Of the options shown, big-east (or beast) is used for HYCOM. If a storm exceeds this domain, the stand-alone HWRF is run. The HYCOM configuration includes a curvilinear mesh, meaning observations are more dense further west. Vertical resolution is 26 hybrid layers and matches RTOFS while horizontal resolution is equal to every other point of RTOFS (or 8km to the west varying to 28 km to the east). A leap-frog numerical integration is used, state variables are on a staggered Arakawa-C grid, and initial and boundary conditions are from RTOFS. Next, Dan explained in more detail the HyHWRF script structure. The main difference from the HWRF is the rtofs_init script in HyHWRF, which is for ocean initialization. To run the system, first a sub-domain (big-east) is selected with IC from a HYCOM analysis run and BC from RTOFS or a previous cycle. Next, forcing files are created using the GFS and parametric winds from the latest tcvital file. The forcing and parametric winds are melded to the GFS background wind. Finally, the HYCOM analysis is run. HyHWRF output is in binary format.
Next Yuejian Zhu and Zhan Zhang gave a joint presentation on the HFIP plan for global and regional ensembles. Yuejian first presented the global ensemble work plan beginning with the 2009 high-resolution GEFS (T574L64) demo which was run once per day from September 1-20, 2009. This ensemble used GSI analysis at T382 with an upgrade to T574 and ETR (ensemble transform with rescaling). The ETR was not tuned and was run once per day. Ensemble output included tracks for ensemble members and the ensemble mean. Results from 2009's Fred (07L) show the operational GEFS (with lower resolution) unable to capture Fred (top panels). However, the higher resolution GEFS clearly showed features from Fred (bottom panels). Yuejian concluded his presentation with the 2010 GEFS demo plan. The main differences are GSI transitioning to T574 and an ETR tuned for rescaling and running every 6 hours. The experiments will involve storms from the 2008 hurricane season with ensemble member tracks, ensemble mean and ensemble spread as output.
Zhan Zhang then presented the work plan for regional ensembles. Zhan began by presenting plots of track error and frequency of superior performance for HWRF V2 ensemble runs compared with the current HWRF forecasts. The ensemble (in red) had lower error for Hanna (08L) and Bill (03L) compared to current HWRF (in blue) and a higher frequency of superior performance. Zhan presented his multi-phase work plan next, starting with work for 2010. Phase 1 involves a 21 member ensemble with T190L28 GEFS ETR initial perturbations using HWRF V3.2 at 27/9 km resolution. Phase 1 will be run for Fay, Gustav, Hanna, and Ike from the 2008 Atlantic season. Phase 2 also has 21 ensemble members with 5 T190L28 GEFS ETR members using two PBL schemes and two convective parameter schemes. The version of HWRF, resolution, and test cases are the same as phase 1 but phase 2 includes cluster analysis post processing, to be developed. The next two ensembles, from Zhan's 2011 work plan, use HWRF V3.2 with 27/9/3 km resolution and will be run for a select group of cases from the 2010 hurricane season. The initial condition-based ensemble will involve 21 members with HWRF-consistent ETR perturbations while the physics-based ensembles will also include PBL and convective schemes. Zhan concluded by describing the final three ensembles from the 2011 work plan. The multi-model ensemble will involve collaboration with FSU as well as a Kernel Density Estimation (KDE) technique, and the Stochastic perturbation in physics work will study uncertainties in the model physics package as part of a collaboration with scientists at EMC and ESRL. Finally, the data assimilation ensemble will involve close work with the DA group at EMC to provide a future hybrid DA system for HWRF.