Your premise is incorrect.
Numerical models normally have:
- A full temperature profile
- A specific field for the 2 metre temperature
- A specific field for the skin temperature
For example, see ERA-40 daily fields, which both 2 metre temperature and Skin temperature, or CFS reanalysis 6-hour fields, which has Air temperature, Skin temperature, and Surface air temperature, (I think surface air temperature means 2-metre temperature, but check their full documentation to be sure).
It is not accurate to take the lowest temperature level in the temperature profile for either of the two, nor to interchange skin temperature with 2-m temperature. But, in some cases, the data distribution may be such that either or both of these are actually attached to the temperature profile. For example, the CFS hourly timeseries has the question:
Where is Sea Surface Temperature (SST) in CFSR? Where is Skin Temperature?
CFSR does not label SST and skin temperature grids differently than temperature grids on other levels. Select the parameter 'Temperature' and the level 'Ground or water surface' to obtain the SST/skin temperature grids, and then use the land/sea mask to determine whether a gridpoint is over water (SST) or land (skin temperature).
(I had the opposite problem, needing skin temperature, and being tempted to use 2 metre temperature instead, until I realised where the data were).
If you encounter a model that really only has the full temperature profile, then I would contact the authors to inquire. Maybe it is a model that really only does well in the free atmosphere, and is not suitable for the extraction of (near-) surface temperatures. The boundary layer can be a hard thing to model!