Lead author here. We found further evidence of drag saturation in hurricane winds, which was independently discovered in the field and laboratory in 2003. Drag saturation means that the air-sea drag coefficient ceases to increase with wind speed beyond some value. This happens in strong winds, between tropical storm and Cat. 1 hurricane force. It's one of essential key unknowns for accurate hurricane prediction.
Data from the leading laboratory study that discovered drag saturation (Donelan et al., 2004) was used soon after to implement the surface drag parameterization for tropical cyclones in the most widely used numerical weather prediction model (WRF). The original study is highly cited and referred to by influential review papers. The WRF parameterization in question has been used by many tropical cyclone prediction studies.
During our analysis, we discovered an error in the original Matlab source code used by Donelan et al. (2004). The error appears in the scaling of in situ wind to the reference height of 10 m. With the original stress data, which we have, the correction is straightforward. The corrected data are now in agreement with the field data in low-to-moderate winds, which is encouraging.
The bottom line is, the original data underestimated the level of drag saturation by 34%, and overestimated the wind speed threshold at which the saturation occurs by 12%. If you use any isftcflx setting other than 0, the air-sea drag in your WRF model is significantly underestimated.
1 comment
[ 3.6 ms ] story [ 17.6 ms ] threadData from the leading laboratory study that discovered drag saturation (Donelan et al., 2004) was used soon after to implement the surface drag parameterization for tropical cyclones in the most widely used numerical weather prediction model (WRF). The original study is highly cited and referred to by influential review papers. The WRF parameterization in question has been used by many tropical cyclone prediction studies.
During our analysis, we discovered an error in the original Matlab source code used by Donelan et al. (2004). The error appears in the scaling of in situ wind to the reference height of 10 m. With the original stress data, which we have, the correction is straightforward. The corrected data are now in agreement with the field data in low-to-moderate winds, which is encouraging.
The bottom line is, the original data underestimated the level of drag saturation by 34%, and overestimated the wind speed threshold at which the saturation occurs by 12%. If you use any isftcflx setting other than 0, the air-sea drag in your WRF model is significantly underestimated.