Problem with Skin Temperature
                        ------- ---- ---- -----------


Note: this problem only occurs for the 1986-1993 data.  For the other
years, parameterization was modified so that sensible heat flux would
not vanish under the following conditions (1) over land, (2) daylight
hours, (3) very convectively unstable boundary layer (based on the
Richardson number).  For all other conditions, the parameterization 
was unchanged.  Evidence was shown that this change had little effect
on the analyses (outside of Tmax and Tsfc).


The skin temperature is determined diagnostically.  It is temperature required 
to balance the fluxes at the surface.


Air:
             latent heat flux
             sensible heat flux *
             short wave (up and down)
             long wave (up* and down)

Ground:
	     heat flux into top soil layer *

       * term which depends on the skin temperature.



A problem occurs when the winds are weak (less than 0.75 m/s) and the ground
is being heated by solar radiation.  In such cases, the Richardson number can 
assume an extremely large negative value (very unstable condition) and the 
sensible heat flux parameterization breaks down.  With light wind speeds, the 
thermal exchange coefficient can become near zero which eliminates the 
sensible heat flux.  Consequently the short-wave heating must be balanced by 
the latent heat flux, long wave flux, and the heat flux into the soil.  In 
order to increase these fluxes so that they balance the short-wave heating, 
the skin temperature must increase.  We found some unrealistically high skin 
temperatures.  These high skin temperatures only occurred in a relatively few 
number of points and may only last one time step as the skin temperature has 
no inertia.

The extremely high skin temperatures can affect (1) 2 meter temperature,
(2) 2 meter specific humidity, and (3) Tmax (2 m).  The latter quantity is 
most affected as it records a problem that may only last one time step.  
Monthly means appear to be relatively unaffected (no bull's eyes) because 
of the averaging process.

A simple screening test for instantaneous data (surface TMP, 2 meter TMP,
SPFH) is to reject that data if the 10 m wind speed is less than 0.75 m/s
and the point is over land.  Unfortunately this test rejects much more
good data than bad data (1 bad point rejected for ~5,000 good points
rejected*).  The twice daily 2 meter temperature and specific humidity were
screened using this test.  The monthly means were calculated without screening 
the data.  (IMHO rejecting 5000 good points for 1 bad point was not best.)

---------------

  * an estimate 

Note: the above test was extremely simplistic.  I rejected a point if
the temperature was greater than some large value.  There are bad points 
which have less extreme values.