[gembud] surface temperature objective analysis techniques

HI everyone, new user here so please be patient :-)

I'm trying to create an objective analysis of surface temperatures
across Canada.  My goal is to interpolate the grids between stations as
accurately as possibly while still retaining the same actual observed
values within the grids that the surface stations fall in.

Using the RUC 15km as a first guess, SEARCH=3D50, GAMMA=3D0.05 and five
passes, oabsfc gives me a RMS error of about 0.5C.  Knowing the lat-lons
of the surface stations, how can I force the corresponding grid to be
the actual observed surface temperature?

Also, it's been suggested to me to translate the temperatures down to a
uniform surface level before running the Barnes analysis.  Any ideas on
how to do this?

Any thoughts are much appreciated, thank-you.