If you mean the original 6-hr ncep data, then you would need to bring that data into the same format as our 3-hr data, so that the clm would read the 6-hr data and understand it. You would probably write a fortran, ncl, matlab, idl, or equivalent script to do this. I think that the clm would then know what to do with the different time resolution, although this hasn't really been tested recently... In general you may find it useful to go over and understand the parts of clm that read and handle the atmosphere data.
I wonder if a simpler way of approaching this is to create files with 6-hr data by sub-sampling or interpolating our 3-hr files. Again you would have to write a fortran, ncl, matlab, idl, or equivalent script to do this. And the clm should know what to do with the different resolution, but, if not, you would need to go inside the code that reads and handles the atm data to figure out what to do.
I hope you find my suggestions helpful.
Good luck,
Sam
