9.6 Turbulent Kinetic Energy




If you are not continuing from the previous section, first Reset, then retrieve conc_case_control.txt and conc_case_setup.txt. In the previous sections, we estimated the strength of the mixing based upon various stability parameters available from the WRF meteorological model output. However some models, such as the NARR, provide a direct estimate of the mixing through the Turbulent Kinetic Energy (TKE) field.

  1. The WRF configuration used in the previous examples did not have the TKE field. Therefore, open the setup menu and change the meteorological data file from captex2_wrf27.bin to captex2_narr.bin. Then save and re-run the simulation. The NARR plume, although not extending as far downwind, looks similar to the WRF simulation result. The HYSPLIT default settings are to use fluxes for stability and the Kanthar-Clayson equations for mixing. When using NARR data, HYSPLIT automatically switches to use temperature profiles for stability and TKE for mixing because the momentum flux is not available in the NARR data and the NARR TKE field has been shown to provide better results for a variety of different simulations. These automatic changes are noted by an obscure line in the MESSAGE file indicating the use of the TKE field (KBLT=3).

  2. For the dispersion calculations the turbulence components are derived from the definition of the TKE
      E = 0.5 (u’2 + v’2 + w’2)

    and the relationship between the horizontal and vertical components is either defined internally by turbulence equations or explicitly by the user through the turbulence anisotropy factor:
      w’2 / (u’2 + v’2).

  3. The Kanthar-Clayson equations are the default for all input meteorological data except NARR. To use these equations rather than TKE for mixing, the option must be selected in the menu. Open Menu #7, press Reset, and select the Kanthar-Clayson radio-button for vertical turbulence and Computed from U/T profile for stability. Save the changes and run the model. Comparing the NARR plume with the WRF plume shows that the NARR concentrations are lower, but the timing and shape of the plumes is comparable. In this case, neither used the TKE field and both used the same turbulence and stability schemes. The only difference between the two calculations was the underlying meteorological data, WRF versus NARR.

  4. Before re-running the TKE calculation with NARR, explicitly select the TKE radio-button. The TKE field only provides information about the total turbulence. The turbulence partition between the vertical and horizontal components must be defined (the anisotropy ratio). The default calculation assumes a value of 0.18. By selecting None in the turbulence anisotropy factors section of the Menu #7, HYSPLIT will compute the factors using the Kanthar-Clayson equations. Save the changes and run the model. The result shows a plume that seems more similar, in terms of timing and peak concentration, to the WRF calculation that did not use the TKE. Note that when the anisotropy ratio is fixed, the mixed layer depth is not required because the mixing is defined at all levels. However, if the anisotropy factors are set to zero, the model will use the turbulence equations to compute the anisotropy factors within the PBL and default assumption (vertical=horizontal) above the PBL height.

  5. To test the sensitivity of the calculation to the anisotropy ratio, open Menu #7 again and change the factors to 0.05, an extreme change to illustrate its effect upon the calculation. Save all menus and run the model. The resulting plume graphic shows a slightly more circular plume and with higher concentrations, the result of less vertical mixing.

Some meteorological data have turbulence values already prescribed and they do not need to be diagnosed from the stability values. The comparison between the NARR and WRF shows the sensitivity of the results to both the underlying model data and mixing methods employed. The main purpose is to illustrate that there may be multiple solutions, all of which are equally correct. The issue of how to deal with an ensemble of results will be addressed in a later section.