I have encountered some sort of bug in WRF 3.0.1.1 in the Noah LSM
urban canopy model.  

The skin temperature (TSK) over Seattle,WA in our real-time 12-km
45-hour forecast valid for 9Z 20 Nov 2008 shows a temperature of 263.2
Kelvin, that's 14 F!  Needless to say the current ground temperatures
in Seattle are closer to 45 F.

Taking a look at the following loop:

http://www.atmos.washington.edu/~ovens/loops/wxloop.cgi?/home/user_www/ovens/urbanbug/plots+all

I think you'll be able to see that wherever the land-use is urban
(the gray color in the last frame of the loop), there are some
seriously cold temperatures especially at forecast hours 42 and 45.
The image titles mean:
 WRF-GFS no ucm 12km Domain  -- ucmcall is set to 0, everything ok
 UW WRF-GFS 12km Domain      -- ucmcall is set to 1, way too cold urban areas

In
http://www.atmos.washington.edu/~ovens/urbanbug/
I have placed the namelist.input file, the RIP input files used to make the plots, and
the gzipped versions of the wrfinput, wrffdda, and wrfbdy files used
for this simulation.