MAP - 

For anyone in the CONUS following the medium range forecast over the past ~8 days, 
you may have had the feeling that the models and ensembles did not handle the 
forecast too well - and you'd be right. Attached are five images that illustrate a 
rather "impressive" failure on the part of both the GFS and ECMWF ensembles:

1. Spaghetti verification (564dm) from the 204 hr forecast valid 12Z 12 Nov 
(GFS/ECMWF/CMC ensemble members, their means, and the deterministic runs).

2. "Outside the envelope" graphic (204 hr fcst valid 12Z 12 Nov.) showing 
the areas where the verifying height lines fell outside the GEFS members' spread.
 
3. As in 2., except for the ECMWF ensembles.

4. Verifying 500mb height pattern with standardized anomalies 
(courtesy Mike Bodner / WPC) showing a strong +3 anomalous ridge over the 
Aleutians and a -5 anomalous upper low north of Hawaii, with a robust ridge/trough over the CONUS.

5. ECMWF ensemble normalised standard deviation from their public website 
(192 hr forecast valid 12Z 12 Nov... 204 hr forecast not available to the public).

Interesting to note how the ridge forecast just south of the 
Aleutians was within the ensemble spread (img 2,3). Although there were many 
members from both ensemble systems (img 1) that indicated energy would tuck back 
southwestward under the ridge in the Pacific... thereby holding back a wholesale 
trough that was forecast to move into the West Coast... there are zero members that 
show strong ridging along 110W and the troughing moving through the Great Lakes. 
The green shading near the OH valley in the EC ensembles (img 5) is too often interpreted 
as "above average confidence" in the forecast, which was misleading in this example. 
Is there another way of indicating uncertainty that is not related to the current ensemble spread 
[d(ensemble spread)/dt?]. Was this a missed opportunity for a "seasoned" forecaster who 
recognized the pattern and its possible outcomes NOT forecast by the ensembles? Or are we 
falling into a trap of thinking the ensembles have all the answers, when we know that they don't?

Lastly, this was not just a "bad" ensemble run. The poor forecasts continued 
in both the GEFS/ECENS for several more cycles (~5 or so) before the shift in the forecast was evident.

Tony