July 28, 2011 Meeting Summary
In the first meeting presentation, Dr. Il Ju Moon from Jeju National University talked about "A Physics-Based Parameterization of Air-Sea Momentum Flux at High Wind Speeds and Its Impact on Atmospheric and Oceanic Modeling." Dr. Moon first described the motivation for this work was to study air-sea momentum fluxes because they provide the boundary conditions for atmospheric and ocean models. The air-sea momentum flux is also a function of the drag coefficient (Cd), which is related to the roughness length (z0). A non-dimensional roughness length widely used in atmospheric models is the Charnock Coefficient. Dr. Moon next detailed different methods for estimating air-sea momentum fluxes. First, the bulk formula can be used with the Cd being a function of wind speed and the Charnock Coefficient being held constant. At low wind speeds, this method shows a monotonic increase in Cd, but this method would have to be extrapolated for higher wind speeds. The second method involves finding the relationship between the Charnock Coefficient and the wave age (which is a function of phase speed and the friction velocity). A small wave age means a young, growing sea. A large wave age means an old sea. Dr. Moon explained that several relationships are proposed based on different observational data. Most ocean models show Cd increasing with increasing wind speed, like in WaveWatch III (WW3). This increase in Cd is unrealistic based on recent studies which show Cd ceases to increase with increasing winds. This is shown in recent experiments conducted using the Coupled Wave-Wind (CWW) model by Moon et al. which feature Cd leveling off at higher wind speeds.
Next, Dr. Moon discussed the CWW experiments, which involved ten hurricanes. Using Isabel as an example, Dr. Moon showed the asymmetric distribution of the storm with the largest drag on the right side of the track. In a scatterplot comparing the Charnock Coefficient and wave age, Dr. Moon showed the dependence on wind speed. Lower winds showed young waves producing higher Cd while higher winds showed young waves producing lower Cd. Thus, Cd levels off at higher wind speeds. A plot was also shown on how these CWW results affect the spatial distribution of a storm. The asymmetric drag distribution shown with Isabel cannot be reproduced using the drag formula, thus coupling to a wave model is important.
Dr. Moon then explained the use of a Coupled Hurricane-Wave-Ocean (HWO) model based on the operational GFDL-POM coupled model from 2006. The HWO model was used to investigate hurricane intensity and structure prediction. Comparing the operational GFDL-POM and the coupled HWO, Dr. Moon showed (for Hurricane Ivan, 2004) Cd leveling off for HWO at higher wind speeds and Cd increasing in the operational model. The spatial distribution and surface winds showed an asymmetric structure in the coupled HWO model, which better agreed with observations. A more symmetric distribution was seen in the operational model. In track and intensity plots, the tracks for both the operational and coupled models are in close agreement with coupled model intensity improved.
Next, the HWO model was used to investigate hurricane wave simulations. Dr. Moon explained that the CWW uses a different parameterization in WW3 than the original. In the original parameterization, young waves led to higher Cd which resulted in good performance. This was most likely due to an overestimation of winds by WW3 originally. When the original parameterization is compared to more accurate Cd values, WW3 is shown to overestimate the significant wave height. More accurate Cd also yields a lower MAE and RMSE value compared to the original. The effect of wave coupling on storm surge modeling was the next topic of Dr. Moon's presentation. In the experiments he showed, Typhoon Maemi was simulated over three domains (coarse, high and very high resolution) in the KORDI-S storm surge model. Three types of wind-stress parameterizations were used: Wu's linear, WW3's fast increasing Cd, and CWW's leveled off Cd. Overall, a higher surge was seen using higher resolution. At higher resolution, the CWW parameterization showed the most accurate surge height.
Dr. Moon then described the empirical method for determining z0, when wave coupling is not available. Here, z0 is a function of wind speed and sea state dependence is ignored, and these values of z0 are based off of the model simulations from 10 hurricanes previously mentioned. In this empirical method, Cd monotonically increases with wind speed when W is less than or equal to 12.5 m/s. When W is greater than 12.5 m/s, Cd levels off between 2 and 3. When this z0 value was tested in eleven cases, the skill of the empirical parameterization was comparable to the full wave-coupled model but the hurricane structure was not accurately represented because of a lack of asymmetry.
Dr. Moon concluded by presenting a summary which stated the main points of his presentation. First, air-sea momentum flux is very wave-dependent. Second, coupled wave-wind model simulations show that Cd ceases to increase at higher wind speeds. Third, coupled hurricane-wave-ocean model experiments demonstrate that hurricane intensity and wind structure and storm surge prediction can be improved by wave coupling. Finally, if a wave model is not available an empirical parameterization can be used. Dr. Moon then briefly discussed some of his future work.
Young Kwon gave the second presentation of the meeting which showed some preliminary results from recent HWRF bug fixes, deep convection tuning, and new PBL/shallow convection schemes. Young began his presentation by describing the recent bug fixes in order of largest impact to smallest. They included an SAS deep convection bug fix involving out-of-bounds moist static energy, nest motion bug fix involving the initialization of fractions of hydrometeo species in start_domain_nmm.f, microphysics bugs, and carbon dioxide transmission data bugs involving a lookup table problem. To counteract negative effects of bug fixes, Young added deep convection tuning. This tuning modified the momentum mixing from 0.55 to 0.2. Plots showing track and intensity illustrated why a deep convection tune-up was necessary. Next, Young presented results comparing the new version of HWRF (with bug fixes, tuning, and new PBL/shallow convection schemes) to an older version of HWRF for Hurricane Dora 2011. Track error was greatly reduced and the northward track bias was eliminated. Intensity errors were not reduced but not considerably worse. Overall, Young demonstrated promising results.