Snow
                                   ----

The snow coverage was estimated from satellite imagery (once per week) 
and snow coverage was converted into a snow depth by assuming it is a
simple function of the surface temperature. (Yes, it is unrealistic but 
we have no snow depth observations.)  For the purposes of data assimilation, 
the snow depth is not critical.  Radiative properties depend on the coverage 
rather than depth.  In addition, if the snow coverage happened to melt away 
(because the snow wasn't deep enough), the snow coverage gets reinitialized 
with "observations" every six hours.

It is suggested that snow coverage be used rather than the snow depth;
i.e., convert the snow depth (WEASD) into 1's or 0's.

The skin temperature over a snow surface is determined by a balance
of the fluxes.  

    Net solar + net long wave + sensible + latent + 
          conduction through the snow layer + snow melt = 0.  

The snow melt term is set to the minimum amount required to keep the
the skin temperature at or below freezing.

Within the snow layer, the temperature is assumed to vary linearly
between the skin temperature and soil temperature (top layer).  If 
the soil temperature is above freezing, there will be snow melt 
which will reduce the soil temperature.  (Of course, the snow depth 
is suspect.)  

OLD:

In Nov/Dec 1999, the NESDIS weekly snow cover was discontiued
and we started using the daily US Air Force snow cover fields.  
This field is global, daily and of a higher resolution.  You
will now see Southern Hemisphere variablility (previously
an older climatology was used).  Since the Air Force product is
a higher resolution, you will see observed snow that is 
incompatible with the model orography.  This can result
in (1) assimilation adds snow (at night?), (2) model melts snow
(during the day?) Repeat steps (1) and (2).


NEW:

From examination, the USAF snow was turned on ~OCT 1998 in
the reprocessed CDAS.

NO SNOW: 1973- start of NESDIS snow