May 14, 2009 Meeting Summary
Young Kwon presented a follow-up to his previous presentation titled "Sensitivity of Air-Sea Exchange Coefficients on Hurricane Size and Intensity: Part II - Code Structure and Preliminary Results". In his previous presentation, Young outlined steps he planned to take to improve HWRF intensity forecasts. This included fixing the wind-pressure relationship of the HWRF model and its tendency to produce larger-sized storms with increasing forecast time, which could be accomplished by tuning the air-sea exchange coefficients (Cd and Ch). Sensitivity experiments could also be performed to assess how different values of Cd and Ch affected hurricane size and intensity, and finally, the effects of sea spray could be included once sensitivity tests had been concluded.
Within the code structure of HWRF surface physics, Cd and Ch are interrelated to one another in the subroutine MFLUX2. Thus, the variables related to Cd and Ch should be changed in a manner consistent to any changes made to Cd and Ch. For example, if Cd is changed, tau(x), tau(y), and U* should also be changed. Using the artificial Cd wind speed profile created for the sensitivity tests, Young showed some preliminary results for Hanna 2008083012 (where Ch was not changed). In his plots, H29N is H209, HLIN uses the linear Cd profile, HDEC uses the decreasing Cd profile, and HINC uses the increasing Cd profile. While the tracks for each of these runs are quite similar, HDEC showed intensity forecast values close to that observed. HDEC also produced a storm size similar to H209 and had the closest MSLP and 850 mb wind speed to H209 as well.
For his future work, Young plans to conduct more experiments with more Cd profiles and cycled runs as well as complete a cycled run with new values of Cd and Ch. Once the sensitivity of the hurricane to Cd is assessed, an examination of Ch sensitivity with a fixed value for Cd will be performed. Finally, Young plans to find the optimum combination of Cd and Ch values to produce the most accurate HWRF intensity and track forecasts. For his future work, Young showed a new Cd wind speed profile, which included operational HWRF (opr). It was also suggested that Young use Powell's Cd values in his tests.
Finally, Young mentioned the milestones of the HFIP Physics Team. These included implementing a prototype version of sea-spray parameterization in the operational HWRF, if this is shown to improve forecast skill, and determining the optimal combination of Cd and Ch to produce improved HWRF intensity forecasts and more accurate storm size while maintaining track forecast skill. Another milestone was to exchange surface physics packages among HFIP teams for testing within operational models. The final milestone was compiling a database of observed microphysical and boundary layer parameters from past airborne datasets to compare with simulations coordinated with the diagnostic team.
Yihua Wu was the second meeting presenter, and he gave a preview of his talk for an upcoming AMS NWP conference. Yihua's presentation was titled "Developing Hurricane-Related Inland Flooding Forecasts at EMC". Since the talk will be used at an upcoming conference, it will not be available on the website at this time.
The motivation for Yihua's work was that freshwater inland flooding accounts for a majority of hurricane-related deaths in the U.S. Could existing models/tools be used to provide forecasts for such flooding? As Yihua mentioned, a land surface model (LSM) is the key for predicting inland flooding because it is driven by surface runoff and baseflow. The current LSM within the HWRF is the GFDL slab model. This model has one soil layer, fixed soil wetness and no surface or subsurface runoff. The Noah LSM has several soil layers, varying soil wetness and includes surface and subsurface runoff.
For his work, Yihua conducted two experiments for the following storms: Ike, Gustav, Hanna, Fay, Dean, Rita, Katrina, and Dennis. The two experiments used the GFDL surface layer scheme and GFS PBL scheme over land and water, but N883 used the slab LSM while N893 used the Noah LSM. Track and intensity error plots actually show similar track error values for N883 and N893 through 72 hours. Thereafter, N883 has lower track error values. The same is true through 48h for the intensity error.
Yihua also presented his work with Hurricane Katrina. Plots of 12h accumulated rainfall, 12h surface runoff, 12h subsurface runoff, and forecasted steam flow were shown for this storm. Two experiments were run where the wet run used GFS soil moisture and the dry run used NAM soil moisture. NLDAS was taken as the observed. Yihua pointed out that there was a small difference in the wet and dry runs in the accumulated rainfall and surface runoff plots. However, there was a large difference in the WET and DRY runs for subsurface runoff showing its sensitivity to initial soil moisture. Yihua's work for this is ongoing, and more results will be presented in the future.