I have a pressure outlet at the top of the tank. The only driving force in the model is buoyancy force. The tank is subjected to ambient conditions like sun radiation, amb temp changes etc.

From Ra number calc I know that the conditions in the headspace of the tank are turbulent - I am using std. k epsilon with enhanced wall treatment.

Question: What will be a good value of turbulent intenstity to give at hte pressure outlet ? 4% is good ? I know that from this value I need to calc the initialization values for k using k = 1.5*(I*uavg)^2

please advise.

If you don't have experiments, you may do the sensitivity study on the turbulent intensity at the outlet. Generally, the turbulent intensity of the backflow won't affect a lot.

We've look at variations from 2% to 5%, and found no real change in the results for similar problems.