A. Why would you want to degrade the 4 km data back to a 12 km grid? I understand if its for computational reasons, but why not just run a 4 km grid over the area of interest?
B. I think there is an option to create a 2nd domain with the specified input data, rather than re-interpolating from the 12 km grid.
This requires that both domain 1 and 2 start at the initial time. Otherwise, in order for the solution and LBC's to be consistent you have to interpolate from the coarse grid.
So if everything is equal, you might as well just run a 4 km grid only. Besides, the use of cumulus scheme on the 12 km grid has some potential negative impacts on the inner nest.
The views expressed in this message do not necessarily reflect those of NOAA or the National Weather Service or the University of Oklahoma.
James Correia, Jr