Aloha Ming Chen!
I recently learned that I can use the nco operator ncra to average netCDF files. As a preliminary test, I plan to first average reanalysis data files by first converting them from grib format into netCDF format using cdo operators. It seems that one of the most obvious problems will be the spin up. Although the spin up of a six month long wrf simulation that I perform now only amounts to about 0.14% of the simulation time (6hrs/4320hrs), if I were to average 6 months of reanalysis data then the percentage of spin up would jump concomitantly to about 0.83% (6hrs/720hrs). If I were to average 50 years of reanalysis into one year, the spin up would comprise about 3.5%. A while ago, when I was planning to remove the spin up from my 6-month long simulations, I would include an extra day preceding the time period of interest, e.g., Oct31. I then planned to cut it out, but never figured out how to do it. Do you have any ideas about how to remove spin up from a run?
It's interesting that you mentioned using average soil data as input b/c I was thinking but what if the time period simulated included an extended drought? Thanks too for describing how the air and soil temperatures come into equilibrium during spin up; I never knew that.
Mahalo!