STORM2K

I know it can't all be put in "by hand" so how is it done? Of course there has to be some human interaction for these to be run and that brings to mind the old adage "trash in-trash out". Is there much chance of a "trash in-trash out" scenarion happening?

There are several types of systems that are used to incorporate data into models (3DVAR, 4DVAR). Essentially, the for each cycle, the model uses its previous run's forecast as a "first guess" for an initialization. Upper air data are QC'd by NCEP upon arrival from the WFOs. The obviously bad data is tossed or "corrected". I'm not sure to what extent surface data is QC'd but I'm sure most of that is done by computer. This QC'd data is then input into the model, fitting it into the first guess. That was way oversimplified but if you want to know more about the process, I suggest you google it as it can be very complicated.

TC data is actually entered by hand in some cases. The "Bogus" (or sometimes called Compute) is submitted into NOGAPS, GFS and GFDL/N after a hand analysis. Central pressure, past motion, Radius of Max winds, Outermost Closed Isobar, Radius Outermost Closed Isobar, Eye Diameter, 34kt wind radii, etc are all submitted by hand and then the model either creates a tropical cyclone of those parameters or "nudges" the existing cyclone in the model fields, depending on the model. The models are designed such that if something is way out to lunch and would cause an instability, it is rejected after automatic QC.

Other models, such as UK Met Office and Japan Global Spectral Model use the WMO bulletins issued by the RSMC and automatically parse out data to use as the Bogus (central pressure, wind radii, motion).