Buoyancy effects for turbulence model

In nitf (non-isothermal flow), the way for setting the Buoyancy force is only can set the volume force(-rho*g)?
However in somecase it is not accuracy for compare to the experiment result (the buoyancy effect is low).

I know some CFD software has own defluat Buoyancy model in turbulence model or in laminar model , how about in comsol?

Any comparsion for comsol and experiment for buoyancy has done to know the accuracy?