A decade + of the Eta performance, including that beyond two days: any lessons for the road ahead?
Fedor Mesinger
NCEP/EMC
Abstract:
First tests with the Eta Model that included a comprehensive physics package were done just about
15 years ago. Since that time the Eta has faced a number of challenges in the form of comparison
with results of other models. In addition, different Eta setups in terms of domains size and
resolution have been run for extensive periods. Some of the ensuing comparison results are
recalled as well as presented, and commented upon. As to more recent times, an effort is mounted
to assess the performance of the Eta beyond two days, in view of the extension of its forecast
time to 84 h in April 2001. Given that in its operational setup the Eta is using the lateral
boundary condition of the Avn run initialized 6 h earlier, one might expect that beyond about two
days the Eta skill falls behind that of the Avn of the same initial time, due to the inflow of the
less accurate lateral boundary data. Three kinds of statistics are inspected aimed at identifying
this anticipated loss of skill: Eta vs Avn precipitation scores at later vs those at earlier
forecast times, rms fits to raobs as a function of time, and accuracy in placing the centers of
major lows at 60 h forecast time. In none of those sings of the Eta relative loss of skill were
found evident. Possible implications are commented upon, in view of guidance they may offer in
attempts to improve NWP skill still further in the years to come.