Jump to content
  • Member Statistics

    17,600
    Total Members
    7,904
    Most Online
    ArlyDude
    Newest Member
    ArlyDude
    Joined

Blizzard Ensemble Verification


Recommended Posts

Attached is a map of ensemble mean sea-level pressure low positions from the 00Z runs

starting December 19th and ending December 25th, verification on Dec 27th at 00Z.

Each point represents the forecast from either the mean NAEFS or the mean ECMWF ensemble.

For verification purposes, the actual position of the of the low is plotted with the

red icon at 00Z on the 27th.

post-2513-0-96318700-1293405698.jpg

Points I think are reasonable from looking at the map:

NAEFS:

1. The NAEFS did relatively poorly, forecasting a low too far offshore. The only 00Z

ensemble that verified better than the ECMWF ENS was at 120 hours, which was a run

where the ECMWF flip-flopped. It was not until the 48 hour forecast (00Z 25th ensemble

output) where the NAEFS placement and low intensity verified well.

ECMWF:

1. It did considerably better than the NAEFS, however hours 168 & 192 (19th and 20th 00Z

output) were considerably too far offshore and too progressive. However, by hour

144 (21st at 00Z) the low placement verified well. At hour 120 (22nd at 00Z) the model

flip flopped and this was the only 00Z run where the NAEFS showed skill vs. the ECMWF ENS.

Starting on the 22nd at 00Z and continuing to the 25th the placement of the low seemed

quite reasonable. The position of the low at hr 48 (25th/00Z) and hr 144 (21st/00Z)

were about the same.

Even with all the operational runs flip flopping, I think the ECMWF ENS performed

adequately aside from the 120 hr flip-flop. The ensembles in general showed much

more consistency than the operational runs.

Comments? It would be interesting to hear post-storm thoughts on last week's forecasts,

especially considering stuff like this:

HPC Dec 24th

INITIALIZATION ERRORS IN NUMEROUS DIAGNOSTIC

QUANTITIES...INCLUDING HEIGHT/VORTICITY FIELDS/RH...ARE EVIDENT IN

BOTH THE 12Z NAM/GFS WITH SMALL BUT LIKELY SIGNIFICANT SHORTWAVE

TROUGHS OVER SOUTH DAKOTA/NEBRASKA ALONG WITH

SASKATCHEWAN/MANITOBA...WITH THESE AREAS ALSO NOT PARTICULARLY

RESOLVED OR PREDICTED WELL BY THE 00Z ECMWF. THUS...THE SPECIFIC

PREDICTIONS BY ALL DETERMINISTIC GUIDANCE ARE IN QUESTION...WITH

THE RECOMMENDATION TO FOLLOW CONTINUITY...WITH THE FINAL OUTCOME

MOST BELIEVED TO LIE BETWEEN THE 06Z GFS AND 00Z ECMWF...WITH ALL

ENSEMBLE GUIDANCE INCLUDING THE SREF MEAN/GEFS MEAN (EXCEPT NOT

THE 12Z VERSION)/ECMWF ENSEMBLE MEAN ALSO CONSIDERED USEFUL TO

ADDRESS THE CONTINUED UNCERTAINTY. THIS APPROACH DISREGARDS THE

SUBSTANTIALLY DEEPER AND WESTWARD SHIFT OF THE 12Z GFS REGARDING

THE POWERFUL LOW TRACKING UP THE EASTERN SEABOARD...AND TO A

LESSER EXTENT THE 12Z NAM WHICH LIES NEAR THE FAST EDGE OF THE

GUIDANCE WITH THE DEVELOPING LOW.

Link to comment
Share on other sites

Excellent analysis.. Could the eastern bias of the NAEFS have been overly influenced by the GGEM which continually showed a more eastern track?

Yes. I would have included GENS data as well but unfortunately I couldn't find the retrospective maps online. If anyone wants to do that using the above map feel free!

As far as models getting lucky, I find the initialization and parametrization process so complex I can't determine if a model verified, but for the wrong reasons. Personally, I question when someone determines a model didn't initialize correctly. Obviously the models guesstimate atmospheric processes more often than not, however models can still provide useful information even though they aren't properly initialized. I'm open to comments from more educated meteorologists than I, but I certainly can't determine why a model verifies or doesn't. For example, I'd guess the NAM has the problems it does because of the mesoscale parametrization processes and perhaps spin-up gone rogue, but that is just an uneducated guess.

Link to comment
Share on other sites

It seems to my unscientific eyes and ears that this was the most poorly forecasted major storm in the last few years. Even as late as last night the forecast for MBY, 10 mi S of DC, was about 6" with 4" much further west in this county. As of the end of the storm, now, we've had a thick dusting, maybe 1/4" (haven't measured). Can anyone identify the reason for the frequent flip-flops and major errors, even less that 12 hours before the onset of the snow?

. At hour 120 (22nd at 00Z) the model

flip flopped and this was the only 00Z run where the NAEFS showed skill vs. the ECMWF ENS.

Even with all the operational runs flip flopping, I think the ECMWF ENS performed

adequately aside from the 120 hr flip-flop. The ensembles in general showed much

more consistency than the operational runs.

HPC Dec 24th

INITIALIZATION ERRORS IN NUMEROUS DIAGNOSTIC

QUANTITIES...INCLUDING HEIGHT/VORTICITY FIELDS/RH...ARE EVIDENT IN

BOTH THE 12Z NAM/GFS WITH SMALL BUT LIKELY SIGNIFICANT SHORTWAVE

TROUGHS OVER SOUTH DAKOTA/NEBRASKA ALONG WITH

SASKATCHEWAN/MANITOBA...WITH THESE AREAS ALSO NOT PARTICULARLY

RESOLVED OR PREDICTED WELL BY THE 00Z ECMWF. THUS...THE SPECIFIC

PREDICTIONS BY ALL DETERMINISTIC GUIDANCE ARE IN QUESTION...WITH

THE RECOMMENDATION TO FOLLOW CONTINUITY

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...