Precipitation #1
                            ------------- --


The Reanalysis precipitation fields are accumulations of short-range
(6-hour) forecasts from the NCEP T-62 Reanalysis model.  Separate forecasts 
of total rainfall (PRATE) and convective-only rainfall (CPRAT) are
provided, thus "large-scale" (stratiform) rainfall can be obtained by 
subtracting the convective-only portion from the total (PRATE-CPRAT).
The units are kg/m**2/s, and "mm/day" units can be obtained by multiplying 
the values by 86400.

Monthly means of these precipitation forecasts generally compare well with 
rain gauge analyses over the Northern Hemisphere land areas.  In the
tropics, comparisons between model rainfall forecasts and satellite
rainfall estimates indicated that the former depict the monthly mean
features reasonably well, but that the intra-monthly and intra-seasonal 
variability is poorly represented in the model forecasts.

Actual rainfall amounts forecast by the model are generally thought less 
useful than differences between the forecasts.
....................................................
.   John E. Janowiak                               .
.   Climate Prediction Center/NCEP/NWS/NOAA        .
.   Phone: (301) 763-8227                          .
.   FAX  : (301) 763-8395                          .
.   Internet Address: wd52jj@sun1.wwb.noaa.gov     .
....................................................



         


                            Precipitation #2
                            ------------- --


The precipitation is heavily based on the physical parameterizations in the 
model and does not use rainfall observations, radar or satellite estimates 
of rainfall.  The precipitation is based on the 0 to 6 hour forecasts 
started from the analysis.  You must remember that the clouds and 
precipitation are not analyzed but are predicted from the winds, temperature 
and humidity using the model parameterizations.

We've noticed systematic errors between the model's predicted precipitation
and observations.  Especially notable is the excess rainfall in the
South-East United States during the warm season.  Tests, however, suggest 
that the precipitation anomaly (annual cycle removed) is more reasonable.

Stationary features related to the Gibb's effect on the model elevation
have also been found.


                            Precipitation #3
                            ------------- --


As always, the Reanalysis precipitation is based on the model physics and
is not based not rain gauge, radar or satellite estimates of precipitation.
In general, model parameterizations of clouds and precipitation are often
considered to be two of the weakest areas in atmospheric modeling.
(This situation may change with high-resolution models which can use 
"micro-physics".)  For people who need the actual precipitation, rain gauge 
data is the most accurate, followed by radar and then satellite estimates 
of precipitation.  The Party line is that the anomalies in the Reanalysis
precipitation are more similar to the observed anomalies than the Reanalysis 
precipitation is to the total rainfall.



See SPURIOUS.H2O
See SPURRAIN.PS (postscript file)
See XPRATE