Large Eddy Simulation on Entire Tropical Cyclones

Turbulence in tropical cyclone boundary layer causes gusts near the surface, and also affects the cyclone-scale structure and dynamics through transports of momentum and sensible/latent heat. Most numerical models of a tropical cyclone do not have a resolution to explicitly simulate turbulent eddies in the boundary layer, and their effects are parameterized. However, such parameterizations introduce significant uncertainties.

Thanks to the K computer, which is the Japanese most powerful supercomputer, we are now able to perform a large eddy simulation (LES) of an entire tropical cyclone, where a horizontal grid spacing is taken to be fine enough to reliably resolve large eddies in the boundary layer. Such an LES contributes to understand roles of micro-scale processes in a tropical cyclone, thus contributing to reduce the uncertainties due to their parameterizations.

In the present study, JMA-NHM (Japan Meteorological Agency's Non-Hydrostatic Model) is used as a LES whose horizontal grid spacing dx is 100 m everywhere. The computational domain covers 2000 km by 2000 km in the horizontal and 23 km in the vertical directions, and horizontal boundary conditions are doubly cyclic. The grid number is 20000 by 20000 in the horizontal directions, and 60 in the vertical direction where grid spacing increases with increasing height. Before starting the LES, a preliminary run with JMA-NHM with dx=2km is made. In this preliminary run, a tropical cyclone develops from a weak initial vortex to a mature stage after 120 hours integration. The grid point values of this mature stage are interpolated to prepare the initial condition for the LES. The time integration of the LES is then performed for 10 hours.

A comparison of the results of the LES with those of the preliminary run at the same instants shows that the minimum surface pressure and the maximum surface winds are nearly the same. However, the radius of the eye-wall is smaller and the radial flow near the surface is stronger in the LES.

Horizontal rolls whose horizontal scale is less than or close to 2 km are found in the boundary layer in the LES. There are two different types of rolls: Type-A rolls occur outside of the radius of the maximum wind (RMW) and have their axis nearly directed in the tangential direction with slightly deflection to the center of the cyclone; Type-B rolls are found inside of the RMW, and have their axis slightly deflected to the outside of the cyclone. Type-A rolls appear to be caused by the inflection point instability of the Ekman layer as in the previous idealized LES studies. They enhance turbulence mixing in the Ekman layer to cause stronger radial inflows, which may have contributed to shrink the radius of the eye-wall in the LES.

On the other hand, Type-B rolls appear to be due to a parallel instability, which has been found in analytic solutions or laboratory experiments of the Ekman layers with strong rotation (i.e. low Rossby number). This instability is possible only near the RMW where large centrifugal force operates. Local maximum surface winds in the tropical cyclone occur at the downdraft regions of the rolls where the momentum is transported downward. Unlike Type-A rolls, the circulations of Type-B rolls do not connect to the wall clouds above.

The LES is also used to evaluate the gust factor, which is defined by a ratio of three-second mean to 1-minute mean wind speed. WMO guideline suggests that it is nearly 1.1 for a tropical cyclone over a sea. The LES results show that the gust factor remains about 1.1 at most of the grid points. However, it reaches about 1.5 near the RMW where Type-B rolls prevail, suggesting that they may cause locally more hazardous winds.