Precipitation Adjustment in the Eta Analysis

                Michael Baldwin      

An 80 km capability has been developed that runs on the smaller Cray
J-90's.  This system has been used to continue work on the new precip
adjustment scheme which includes adjustment of the convective as well as
the grid-scale model precip which works in the following way. For the
convective adjustment, if the observed precip is less than forecast, the
latent heat profile is modified like we usually do by multiplying it by
the fraction obs/forecast precip, and fixing the RH to be the same as
before that latent heat adjustment.  No changes are made to the cloud
water.

The new part is when the observed precip is greater than forecast
(regardless if forecast was greater than or equal to zero), then a
different adjustment is made.  The "first guess" temperature profile is
created (basically a slightly modified moist adiabat through cloud base)
and the "max possible precip" is computed by summing up the latent heat
that would be released if the T profile was set to the "first guess"
values.  This profile is adjusted by multiplying the delta T's by the
fraction obs/MAX precip, and if that fraction is greater than 1., it's
set to 1.  So that means that if the obs precip is greater than the
maximum possible from the convective scheme profile, it will only be set
to the maximum value.  The moisture profile is then determined in the
normal manner using a cloud efficiency that has been modified by the
observed precip rate.  The higher the precip rate, the more moist the
profiles will be, up to an upper limit which has RH's in the 60-70%
range.

The "grid-scale" type adjustment is almost like before, multiplying the
cloud water and latent heat profiles by obs/forecast ppt.  If the
forecast precip is zero, setting the cloud base as the first layer above
ground with RH > 65%, and the cloud top as the first layer above cloud
base with RH < 65%.  If the cloud determined in this way is not deep
enough then  we go back to the old way of picking cloud base and top via
the precip rate.  Then a parabolic profile is created and the cloud
water is bumped up to the minimum value and the RH is bumped up to 90%
and the sub-cloud RH's are also bumped up.  The idea behind this is that
even though the model didn't predict precip, it probably has a better
idea of where the cloud should/will be by looking at the RH profile than
we would come up with by doing a cloud depth vs. precip rate scheme,
which will still be used as a last resort.  When these parallel tests
are reinstated, the accumulated precipitation forecasts can be examined
versus observations and versus operational runs at this site:

          http://sgi62.wwb.noaa.gov:8080/verf/pcpgifs.html