Study on improving tropical cyclone forecasts by improving the way turbulence near the surface is modeled published in the Journal of Geophysical Research

This paper shows that correctly representing the details of processes in model physics schemes can lead to big forecast skill improvement (up to 10 kt).  The large number of cases provide confidence in the results

Summary: The planetary boundary layer (PBL) is the lowest part of the atmosphere, typically within about 1 km of the Earth’s surface. The flow in this part of the atmosphere is chaotic, or turbulent, with lots of random motion, and so computer weather models need special approximations to forecast this flow and its effects.  In fair-weather conditions, the PBL is separated from the atmosphere above by an inversion, or layer where the temperature increases with height. In tropical cyclones (TCs), however, intense turbulence can be generated by the warming in clouds (from water vapor changing to liquid water and ice) and extends all the way up to the the top of the TC in the eyewall and rainbands. We tested a new method for including the effects of these phase changes (such as changing from liquid to solid, and how that affects the vertical temperature change in the eyewall, what we call stability) on predictions of TCs. This correction was used to make the model more realistic in the Hurricane Analysis and Forecast System (HAFS), a new NOAA model for TC prediction.

Important Conclusions:  

  1. Using the old method, HAFS does not correctly capture the vertical structure of thunderstorms in the eyewall. 
  2. The correction substantially improves the model’s skill in predicting the intensity of TCs. Compare the red (modified) and blue (original) lines in Figure 1, showing a 10-kt reduction in intensity bias using the modified scheme. This is a big forecast skill improvement with a large number of cases providing confidence in the results. 
  3. The turbulence in the eyewall plays a pivotal role in initiating a TC’s rapid intensification, by focusing strong thunderstorms near the eyewall. See the comparisons between the model runs and observations in Figure 2, where the modifed scheme has stronger thunderstorms (denoted by the reds in the simulated radar image) in the inner core of the storm, while the original scheme has stronger thunderstorms in the outer part of the storm. Observations showed that the newer scheme, with stronger thunderstorms in the inner eyewall, was correct. 

For more information, contact The full paper can be found at

This work is supported by NOAA/HFIP under Grants NA16NWS4680029 and NA18NWS4680057 and National Science Foundation under Grant AGS-1822238.