Date: Fri, 08 Nov 2002 13:59:48 -0500 From: "Geoff Dimego" Organization: DOC/NOAA/NWS/NCEP/EMC Subject: Eta performance Folks; Last month I was asked by Louis to respond to observations of inconsistent behavior of the Eta Model conveyed by Walt Drag of Taunton, MA. Subsequently, Warren Blier of Monterey, CA also wrote. Here is my response to Warren which I realized today that I had not sent to Walt (sorry). I send it to you all now because NCEP's HPC will be looking at this Etay parallel and may be mentioning it in their discussions. Note also that one of our upgrade cycles was delayed this year due to T-254 and will not get implemented until we have transferred to the new machine. It is running as a 12 km real-time parallel Etax. A description of its changes are in the parallel log file at http://www.emc.ncep.noaa.gov/mmb/mmbpll/etapll/ where you can view real-time runs of the 12km Etax. Here is my response to Warren: Warren; We noticed. I don't have any explanation to offer yet. The quality of the Eta forecasts has been more inconsistent this year than in the past and not just in the West. We have been trying to isolate the cause(s). It isn't a simple matter as you can imagine. In an effort to isolate the 3DVAR analysis contribution to this behavior, we have started a parallel run (32 km EtaY) which is initialized off the GFS (what used to be called the AVN - we are under the gun to call it GFS now) analysis. There are differences between the dump times and data availability/usage between these two analyses and there are differences in the models so you should NOT expect to get the same prediction from the two systems, but we've seen enough cases where the GFS has had a better handle on features that we are looking first in the direction of the 3DVAR. You can follow the progress of the parallel runs at Geoff Manikin's ETAY page (look here first) http://www.emc.ncep.noaa.gov/mmb/mmbpll/etay/ and at Eric Rogers' Eta-32 page (graphics run later than Geoff's page) http://www.emc.ncep.noaa.gov/mmb/mmbpll/etapll32/ Warren Blier wrote: > > Hi Geoff, > > Perhaps you've already noted this, but there was a profound > difference in solution between the AVN and Eta initialized > at 00Z Wed Oct 23. Attached to this e-mail are the > respective 500 mb and precip predictions for 60 hrs (12Z Fri > Oct 25), and the Eta 500 mb analysis at that time. The > large difference between the models was operationally > significant for us, as the incoming system appeared to have > the potential for bringing the first significant precip of > the water year to central California. For our CWA, the Eta > solution was much slower and wetter, while the AVN solution > was faster and drier. Ultimately, as the attached analysis > indicates, the AVN did very well and the Eta very poorly. > Is there some particular reason you're aware of for why the > Eta would have been so far off in this case? Perhaps the > later data-cutoff of the AVN allowed inclusion of some > additional critical data? It's going to be a challenge for > us to issue 2.5 km resolution PoP and QPF grids out 7 days > if the 2 major operational models continue to yield vastly > different solutions in critical forecast situations even on > the 1-3 day time scale! > > Regards, > > Warren > > ------------------------------ > Warren Blier, Ph.D. > Science and Operations Officer > National Weather Service > 21 Grace Hopper Ave, Stop 5 > Monterey, CA 93943-5505 > 831-656-1710 x224 (voice) > 831-656-1747 (FAX) > e-mail: Warren.Blier@noaa.gov Steve/Geoff: Pls respond to Walt Drag. I heard this from several of the forecasters last week, but Walt is the only one to put this in writing. Louis -------- Original Message -------- Subject: ETA reminder Date: Mon, 07 Oct 2002 07:41:08 -0400 From: "Walter Drag" Organization: DOC/NOAA/NWS - National Weather Service To: Louis Uccellini Good morning Louis, I thought that was good AMS conf... interchanges etc. On the ETA... it seems to have regressed here over the last year or so in the NE USA. On two recent occasions including an event for 9/26, the ETA behaved as the NGM... no model qpf despite large composite 1000-500 RH >85 pct and wk omega .. it, the ETA, didn't catch up to the better fcst indications of the GFS until within ~ 12 hrs of T0. Not sure if this is a summer bias... but if Jim Hoke asked you about the ETA... then he's either hearing it too or noticing it himself. Good luck with this! BTW: The GFS is clearly the better QPF model here in the NE USA over the last 9 months or so. Walt ------------- End Forwarded Message -------------