NYC/PHL Dec 26-27 Potential - Part 4

[David Reimer]

It looks like HPC is leaning a bit wetter than not:

http://www.hpc.ncep....qpf/p120i12.gif

That's 12 hours old.

[SmokeEater]

It looks like HPC is leaning a bit wetter than not:

http://www.hpc.ncep....qpf/p120i12.gif

That was issued 12 hours ago.

[NorthShoreWx]

It looks like HPC is leaning a bit wetter than not:

http://www.hpc.ncep....qpf/p120i12.gif

That was issued at 7 this morning.

[Collegestudent11]

That was issued 12 hours ago.

Why cant they just put it in standard time? Sorry about that.

[atownwxwatcher]

As i just posted on the main forum general 18 Z thread ...it looks like the 18 Z UKMET has also slowed down from its 12 Z run

http://moe.met.fsu.edu/tcgengifs/ukm/current/slp10.png

[tornadojay]

not sure if this got posted earlier from HPCs afternoon update, but I'll post it.

IN THE EAST...MODEL SPREAD HAS CONTINUED TO NARROW CONCERNING THE

TRACK AND OTHER DETAILS OF A POTENTIAL WINTER STORM AFFECTING THE

EAST COAST DAYS 3-4...WITH THE MOST NOTICEABLE AND CONSISTENT

MODEL TREND OBSERVED FOR FASTER NORTHEASTWARD ACCELERATION.

EARLIER PREFERENCES WERE FOR A SCENARIO CLOSEST TO THE 06Z GFS

WHICH REPRESENTED AN AVERAGE OF THE MOST WESTWARD 00Z ECMWF AND

MORE EASTWARD 00Z GFS...WITH THE LAST SEVERAL RUNS OF THE

UKMET/CANADIAN SOUTH AND/OR EAST OF AN OTHERWISE NARROWING CLUSTER

OF GUIDANCE AND THUS WERE NOT SERIOUSLY CONSIDERED. DESPITE

IMPROVED CLUSTERING OF SOLUTIONS AND SOME CONSISTENCY IN

TRENDS...IT REMAINS PREMATURE TO PICK A SPECIFIC DETERMINISTIC

SOLUTION WITH HIGH CONFIDENCE. THE TROUGH APPROACHING THE REGION

HAS ONLY RECENTLY ENTERED THE MORE DENSELY POPULATED OBSERVATIONAL

NETWORK...AND THUS IT WILL PROBABLY TAKE ANOTHER 1 OR 2 MODEL RUNS

AT LEAST TO RESOLVE THE MULTIPLE STREAM INTERACTIONS WHICH IS

CRUCIAL TO A SUCCESSFUL FORECAST IN THIS SHORTWAVE PATTERN.

THUS...THE FINAL PRESSURES/FRONTS WERE ONLY NUDGED ABOUT MIDWAY

BETWEEN THE 06-12Z GFS...WHICH CREDITS THE NEW CONSENSUS FOR A

FASTER SOLUTION WHILE ALSO MAINTAINING CONTINUITY AS THE PRECISE

TRACK FORECAST STILL CONTAINS A LARGE AMOUNT OF UNCERTAINTY.

[am19psu]

This is generally correct. There are plots out there, but in terms of forecast skill, all four cycles are pretty much equally skillful (statistically speaking). It's always possible for any individual realization (00z/12z included) to be an outlier.

We have a wealth of observations even at 06z/18z (satellite and satellite based products, surface, aircraft, GPS radio occultation, wind profilers, etc.), so this isn't terribly surprising. However, it's always possible for some individual feature to get sampled/resolved (particularly in getting its vertical structure correct) as it reaches the more dense/accurate/reliable radiosonde network.

I've always found the balloons to be critically important in situations that are very sensitive to initial conditions (e.g. severe wx, tropical cyclone tracks) where a few meters difference can make for large variance downstream. Do you find that to be true?

If so, then the old wives' tale about the off hour runs could be true, but really only for the initial sampling. Now that all of the synoptic players are on the field, so to speak, off hour runs should be just as good as synoptic hour runs because of the way data are assimilated. Right?

[Jet-Phase]

If that whole setup was like 100 - 200 miles west, that would have been golden

Yep...what's amazing is how close this run is to the GFS run on 12/18/10. Shift it 100-200 miles west and it's identical from a run 5 days ago!

[dtk]

I've always found the balloons to be critically important in situations that are very sensitive to initial conditions (e.g. severe wx, tropical cyclone tracks) where a few meters difference can make for large variance downstream. Do you find that to be true?

If so, then the old wives' tale about the off hour runs could be true, but really only for the initial sampling. Now that all of the synoptic players are on the field, so to speak, off hour runs should be just as good as synoptic hour runs because of the way data are assimilated. Right?

I largely agree with this (radiosondes are really the only way we get a very accurate sampling of temps & winds in the vertical....and this can be crucial for severe wx/lapse rates/shear, TC tracks/mean steering, etc.).

However, for the centers that only use a 6 hour (+/- 3 hour) window or less, the direct use of the balloons is technically lost for the off-cycle times (we don't assimilate 00z obs at 06z). This implies that the new observations that are assimilated at those times (non-radiosonde) need to correct the previous cycle's 6h forecast without destroying all the information that was gained by using the high quality sondes. I need to come up with a better way of saying/explaining this...

[Bobby]

Can we mention a west trend now? I think we panicked a little early maybe.

C'mon.... That doesn't happen around here............MUCH

[WeatherFox]

And these words of optimism from the NYC NWS discussion as of 5:08 pm today: ...OVERALL...PLENTY OF ENSEMBLE

MEMBERS STILL FORECASTING A FULL FLEDGED SNOWSTORM.

Me: There is still some chance for a little snow surprise especially on eastern Long Island and even in NYC.

not sure if this got posted earlier from HPCs afternoon update, but I'll post it.

IN THE EAST...MODEL SPREAD HAS CONTINUED TO NARROW CONCERNING THE

TRACK AND OTHER DETAILS OF A POTENTIAL WINTER STORM AFFECTING THE

EAST COAST DAYS 3-4...WITH THE MOST NOTICEABLE AND CONSISTENT

MODEL TREND OBSERVED FOR FASTER NORTHEASTWARD ACCELERATION.

EARLIER PREFERENCES WERE FOR A SCENARIO CLOSEST TO THE 06Z GFS

WHICH REPRESENTED AN AVERAGE OF THE MOST WESTWARD 00Z ECMWF AND

MORE EASTWARD 00Z GFS...WITH THE LAST SEVERAL RUNS OF THE

UKMET/CANADIAN SOUTH AND/OR EAST OF AN OTHERWISE NARROWING CLUSTER

OF GUIDANCE AND THUS WERE NOT SERIOUSLY CONSIDERED. DESPITE

IMPROVED CLUSTERING OF SOLUTIONS AND SOME CONSISTENCY IN

TRENDS...IT REMAINS PREMATURE TO PICK A SPECIFIC DETERMINISTIC

SOLUTION WITH HIGH CONFIDENCE. THE TROUGH APPROACHING THE REGION

HAS ONLY RECENTLY ENTERED THE MORE DENSELY POPULATED OBSERVATIONAL

NETWORK...AND THUS IT WILL PROBABLY TAKE ANOTHER 1 OR 2 MODEL RUNS

AT LEAST TO RESOLVE THE MULTIPLE STREAM INTERACTIONS WHICH IS

CRUCIAL TO A SUCCESSFUL FORECAST IN THIS SHORTWAVE PATTERN.

THUS...THE FINAL PRESSURES/FRONTS WERE ONLY NUDGED ABOUT MIDWAY

BETWEEN THE 06-12Z GFS...WHICH CREDITS THE NEW CONSENSUS FOR A

FASTER SOLUTION WHILE ALSO MAINTAINING CONTINUITY AS THE PRECISE

TRACK FORECAST STILL CONTAINS A LARGE AMOUNT OF UNCERTAINTY.

[jm1220]

And these words of optimism from the NYC NWS discussion as of 5:08 pm today: ...OVERALL...PLENTY OF ENSEMBLE

MEMBERS STILL FORECASTING A FULL FLEDGED SNOWSTORM.

Me: There is still some chance for a little snow surprise especially on eastern Long Island and even in NYC.

If I end up with a surprise 3-5" or something, it'll be a lot better than zero.

[Bobby]

Yep...what's amazing is how close this run is to the GFS run on 12/18/10. Shift it 100-200 miles west and it's identical from a run 5 days ago!

Similiar but I wouldn't call anything over 100 miles close. You can have HUGE cutoffs in that space. Waiting on the next run. I still think out here on the island we're in the best shape of anybody save for the Cape area.

[am19psu]

However, for the centers that only use a 6 hour (+/- 3 hour) window or less, the direct use of the balloons is technically lost for the off-cycle times (we don't assimilate 00z obs at 06z). This implies that the new observations that are assimilated at those times (non-radiosonde) need to correct the previous cycle's 6h forecast without destroying all the information that was gained by using the high quality sondes. I need to come up with a better way of saying/explaining this...

Ok, let me ask another question then: aren't obs used to adjust the first guess field from the previous GFS run? IOW, the high quality sonde data would just be iterated 6 hours later. Is that right?

BTW, thanks for taking the time to answer all of these modeling questions. Learning how things actually work in model world ought to dispel some of the erroneous CW and help us all out as forecasters (not just the mets here).

[dtk]

Ok, let me ask another question then: aren't obs used to adjust the first guess field from the previous GFS run? IOW, the high quality sonde data would just be iterated 6 hours later. Is that right?

BTW, thanks for taking the time to answer all of these modeling questions. Learning how things actually work in model world ought to dispel some of the erroneous CW and help us all out as forecasters (not just the mets here).

Right, so 00z sondes would correct a 6 hour forecast from 18z (it's technically not the actual 6hr GFS forecast, but from something initialized in between cycles....we do a catch up cycle in between actual GFS run as more data has trickled in). So this corrected forecast (the analysis/initial conditions) is then evolved out to the next cycle. So the information is still there in the 6z analysis, it's just that we've relied on the NWP model to evolve that information (consistent with the model dynamics) forward in time. So the information extracted from sonde data is still there in model space....it's just been propagated/evolved by the model.

I'm happy to answer questions and thrilled people want to learn how this stuff works (both in general, and NCEP-specific). So long as people have questions/want information, I'll do my best to provide useful replies.

[am19psu]

Right, so 00z sondes would correct a 6 hour forecast from 18z (it's technically not the actual 6hr GFS forecast, but from something initialized in between cycles....we do a catch up cycle in between actual GFS run as more data has trickled in). So this corrected forecast (the analysis/initial conditions) is then evolved out to the next cycle. So the information is still there in the 6z analysis, it's just that we've relied on the NWP model to evolve that information (consistent with the model dynamics) forward in time. So the information extracted from sonde data is still there in model space....it's just been propagated/evolved by the model.

So getting back to the crux of the question, once the s/w (in this case) has made landfall over the observing network, that information will be integrated forward and there shouldn't be much, if any, drop off to the next off hour run?

[tornadojay]

been a while since we've seen these, but here are the 18Z GFS "clown" maps for snowfall amounts..

[earthlight]

At this point 2 inches would be plenty enough for me.

Thats what she said, by the way.

[RowanBrandon]

I'll take 1" at this point.

[ARyan]

Why cant they just put it in standard time? Sorry about that.

Because they forecast for the ENTIRE country.

[dtk]

So getting back to the crux of the question, once the s/w (in this case) has made landfall over the observing network, that information will be integrated forward and there shouldn't be much, if any, drop off to the next off hour run?

Pretty much. It depends on how the model handles the information it's been given (6hr errors aren't huge, but they do exist obviously) and what new information comes in during the subsequent off-hour cycle in the vicinity of the s/w (other obs). Of course, if we could re-sample said s/w with new sondes at the off hour time, we'd be even better off still....

[Collegestudent11]

Right, so 00z sondes would correct a 6 hour forecast from 18z (it's technically not the actual 6hr GFS forecast, but from something initialized in between cycles....we do a catch up cycle in between actual GFS run as more data has trickled in). So this corrected forecast (the analysis/initial conditions) is then evolved out to the next cycle. So the information is still there in the 6z analysis, it's just that we've relied on the NWP model to evolve that information (consistent with the model dynamics) forward in time. So the information extracted from sonde data is still there in model space....it's just been propagated/evolved by the model.

I'm happy to answer questions and thrilled people want to learn how this stuff works (both in general, and NCEP-specific). So long as people have questions/want information, I'll do my best to provide useful replies.

Why do the models run four times a day, compared to other models?

[tornadojay]

At this point 2 inches would be plenty enough for me.

Thats what she said, by the way.

[Rtd208]

Keep your Yule Log in the fireplace my friend lol.

At this point 2 inches would be plenty enough for me.

Thats what she said, by the way.

[ARyan]

Right, so 00z sondes would correct a 6 hour forecast from 18z (it's technically not the actual 6hr GFS forecast, but from something initialized in between cycles....we do a catch up cycle in between actual GFS run as more data has trickled in). So this corrected forecast (the analysis/initial conditions) is then evolved out to the next cycle. So the information is still there in the 6z analysis, it's just that we've relied on the NWP model to evolve that information (consistent with the model dynamics) forward in time. So the information extracted from sonde data is still there in model space....it's just been propagated/evolved by the model.

I'm happy to answer questions and thrilled people want to learn how this stuff works (both in general, and NCEP-specific). So long as people have questions/want information, I'll do my best to provide useful replies.

Sorry...quick clarification question, so you are saying that the "off hour" runs are initialized by modify the soundings based on surface observations?

[atownwxwatcher]

I am really surprised that no one has really commented on how the UKMET @18 Z has slowed down as well.... and how that potentially could lead to some credence of the 18 Z GFS slowing down...

I do not think its out of the realms of possibility that we could be seeing this thing slow down on 00z and perhaps an earlier phase...

All it takes is a stronger Northern Stream energy which was hinted at & a slightly earlier phase.

[famartin]

Why cant they just put it in standard time? Sorry about that.

For one, its a national product and that would mean the time is different everywhere. To be fair, they use a time that all meteorologists understand no matter where you are.

[dtk]

Why do the models run four times a day, compared to other models?

User need and legacy.....unless you're asking why we have a 6 (or 3, or 1) hour assimilation cycle, that has a bit more of a technical response.

[Ridingtime]

User need and legacy.....unless you're asking why we have a 6 (or 3, or 1) hour assimilation cycle, that has a bit more of a technical response.

Question for you as well: Do you believe after this possible non-event the advantages of 4DVAR are negligible and the need for it on the GFS are now presumably not as rushed?

[dtk]

Sorry...quick clarification question, so you are saying that the "off hour" runs are initialized by modify the soundings based on surface observations?

No need to apologize, clarification is good. 6z (18z) runs are initialized by making corrections to the 6 hour forecast from the 0z(12z) runs. The soundings themselves are never modified/used explicitly in the 6z/18z cycles (unless there are special launches valid at those times. However, the information that was put into 00z and 12z initializations exists (with some error) in that 6 hour forecast that is used as the first guess. The only observations used at 6z/18z are those that are valid at those times (+/- 3 hours, things like aircraft, wind profilers, satellite based observations, surface obs., etc.).

Now, if we start talking about 4DVAR with longer time windows, it gets more complicated.....