The following are comments provided by Bill Nichols, lead forecaster at the Davenport, IA

*   Usually, the QPF's are extremely close.  My observations suggest biggest variances
noted with marginal forcing/moisture events and also sometimes in evolution of a
current QPF field.  I've not been able to assess systematic trends in the differences
yet that operationally we can say is an improvement...will keep monitoring.

*   Both model QPF schemes seem to suffer from RH problems in lower levels that
sometimes manifest themselves (or don't) in QPF and Vertical Motion fields.  I find
these RH often overforecasted, and sometimes have major operational impacts.  One
example, last week, several runs of the ETA and ETAX show 80% plus RH that
obviously wasn't going to happen (i.e. center arctic high)...not unusual...and yes, we
were clear all day (DVV).

*   Both schemes suffer from missing or underestimating shallow moisture convergence
with decent low level forcing.  Also...notorious for missing light flurries/sprinkles or
even light showers in late winter and early spring with cold air aloft.

***********LARGEST IMPACT FORECASTING**************************

Operationally, the biggest item all us in our office note (and applies to both schemes)
is the large run to run variance in QPF output.  It is more than any other model and
can also apply to incorrect location(s).  Latest example: 28 January 2003 heavy QPF
band was along Highway 30...yet even 12Z and 18Z runs (both schemes) had heavy
QPF north of Highway 20.  A good 60+ miles error...then there was the east/west
displacement.  If one notes the other models...though not as coherent on the heavy
max strip...they hit the geographic location(s) better of the heavy strip.


Realistically, we heavily merge and rely on conceptual forcing and comparing with
GFS/AVN, NGM (yes, many times the NGM QPF is actually best and normally is
fairly consistent) and UKMET is also very helpful because of its consistency.   This
means many of us often may not even use ETA UVV's with ongoing QPF due to it
being influenced by the ongoing QPF if deemed not viable.  This is our biggest
challenge we routinely face.


If possible...it might be interesting to compute and track QPF standard deviations
compared to other models.  Empirically...I suggest it may show greater variances
compared to the other models all of us routinely note here.   Then, it possibly would
aid in shedding some light to these changes we routinely see with the ETA QPF.
They can...and too often are...very extreme.


In summary, the change to the new QPF fields would be modest at best...and for most
operational forecasts, non sequitur; due to the daily larger scale challenges/biases of
the model.