To get turbulence, you need to run the model in the so-called LES mode. This means:
- very small time and space step (less than 100 m, time step adjusted for CFL condition)
- disabling some physical parametrizations (bl_pbl_opt = 0)
- other stuff
I believe the output fields of WRF are instantaneous fields. But, outputting all fields for every step would make very large files. If you want turbulence spectra, use
tslist. This enables you to get a field at every time step at a particular point of the domain. You can have multiple of these points (I used more than 300 once). Then, you also need to run the model in LES mode.
Finally, just because you have a very small time step (say 0.2 s), this
does not mean that motions at such small scales are resolved. WRF smoothes fields internally, so only motions at scales of ~7 times the time (and space) step are resolved.
EDIT: Sorry, I did not read the last question. WRF field output is instantaneous, but if you are running WRF at a scale which is > 1 km, then this instantaneous output is representative of the 10-min mean output of "real" wind. So no need to average. You only average if you run the model at a much smaller scale.