February storm threat discussion

[Wonderdog]

I have a question about the individual ensemble members. Has there ever been a study as to which of the ensembles, as depicted on the model page, is more accurate than the others?

[chris87]

I have a question about the individual ensemble members. Has there ever been a study as to which of the ensembles, as depicted on the model page, is more accurate than the others?

I believe each individual ensemble member has no statistical relationship to that specific member from a previous cycle. The goal of ensemble forecasting is to quantify the "uncertainty" in model space, assessed by applying perturbations to the initial conditions of each ensemble member [i'm sure dtk can correct any misinformation here!].

[mitchnick]

Bits and pieces from HPC.

THE 00Z ECMWF

IN PARTICULAR COMPARES UNFAVORABLY TO OTHER MODEL/ENSEMBLE

GUIDANCE WITH ITS SHRTWV TRACKING FROM THE SRN ROCKIES INTO THE OH

VLY DURING WED THU......

THE RAPID SHARPENING/AMPLIFICATION OF THE WRN

NOAM RIDGE SEEMS TO FAVOR AT LEAST AN INTERMEDIATE TROF AXIS

VERSUS THE ERN SIDE OF THE ENVELOPE.....

A DIFFICULT AND LOW TO NO

CONFIDENCE FORECAST IN DETAILS LATE PERIOD MAINLY WITH THE ERN

CONUS TROF.....

DEPICTION BY DAY 5 FRI

OF A CENTER JUMPING SFC LOW FROM THE OH VALLEY TO THE NC COAST AND

THAN AN EXIT NEWD TO NEAR THE NEW ENG BENCH MARK BY DAY 6 SAT

SEEMS REASONABLE AT THIS TIME ALBEIT WITH CONTINUED VERY LOW

CONFIDENCE.

we don't know what's gunna' happen, so we'll put a dying slp in OH, pop a new one off the NC coast then move it to the benchmark

ya' think maybe 1 or 2 of the young weenies from Wright Wx have grown up and are now working at HPC? lol

[WinterWxLuvr]

we don't know what's gunna' happen, so we'll put a dying slp in OH, pop a new one off the NC coast then move it to the benchmark

ya' think maybe 1 or 2 of the young weenies from Wright Wx have grown up and are now working at HPC? lol

Regardless of where they came from, I think it's safe to say that they know more than either of us. They did state that it was a very low confidence forecast.

All I did was post what I thought was an interesting write up of the situation for the week ahead.

[chris87]

we don't know what's gunna' happen, so we'll put a dying slp in OH, pop a new one off the NC coast then move it to the benchmark

ya' think maybe 1 or 2 of the young weenies from Wright Wx have grown up and are now working at HPC? lol

they have to make a forecast...they note the uncertainty and present a viable, "albeit, not likely to be accurate" solution.....i'd like to see your discussion..so I can subsequently lol at it, j/k. in general, cut them a break...they seem fully aware of the issues on the table.

[usedtobe]

they have to make a forecast...they note the uncertainty and present a viable, "albeit, not likely to be accurate" solution.....i'd like to see your discussion..so I can subsequently lol at it, j/k. in general, cut them a break...they seem fully aware of the issues on the table.

And neither of the guys writing the discussion are snow weenies. One is nearing retirement and the other is a guy who really knows statistics.

[dtk]

I believe each individual ensemble member has no statistical relationship to that specific member from a previous cycle. The goal of ensemble forecasting is to quantify the "uncertainty" in model space, assessed by applying perturbations to the initial conditions of each ensemble member [i'm sure dtk can correct any misinformation here!].

You ask and you shall receive. First, you're basically correct....and from a scientific (and design) standpoint, every member should be an equally likely outcome.

However, the GEFS system does have one quirk relative to the other major operational centers that can lead to some correlative skill characteristics by "member number" (albeit with a 24 hour lag). The GEFS initial perturbations are applied using something called an ensemble transform (rescaled) method, which for all intents and purposes is a variant of breeding. For each cycle (00, 06, 12, 18), there is actually an 80 member ensemble of short (06h) forecasts. Those 80 forecast perturbations are then rescaled (shrunk) to represent initial condition perturbations (they are also reorthogonalized). From the 80 initial perturbations, only 20 are selected and integrated out as part of the EPS. The problem is that the selection process isn't random:

00 UTC : members 01-20

06 UTC : members 21-40

12 UTC : members 41-60

18 UTC : members 61-80

I guess this was was chosen to try and maintain "forecast continuity". The naming convention for the products does not actually line up with the member number within the perturbation cycling. In other words, member NN" is not the same cycled "member NN" between each of the 4 cycles/day.

I suppose then it's possible through the breeding and selection process there could be relationships between the members (with a 24h lag though, not from cycle to cycle). This has become less likely now that they have implemented stochastic noise in the ensemble model integration.

[mitchnick]

everybody jumps in on my HPC post

obviously I was kidding, but you have to admit that's on he!! of a storm track to pick from out of all the variety of possibilities the models have shown

[chris87]

You ask and you shall receive. First, you're basically correct....and from a scientific (and design) standpoint, every member should be an equally likely outcome.

However, the GEFS system does have one quirk relative to the other major operational centers that can lead to some correlative skill characteristics by "member number" (albeit with a 24 hour lag). The GEFS initial perturbations are applied using something called an ensemble transform (rescaled) method, which for all intents and purposes is a variant of breeding. For each cycle (00, 06, 12, 18), there is actually an 80 member ensemble of short (06h) forecasts. Those 80 forecast perturbations are then rescaled (shrunk) to represent initial condition perturbations (they are also reorthogonalized). From the 80 initial perturbations, only 20 are selected and integrated out as part of the EPS. The problem is that the selection process isn't random:

00 UTC : members 01-20

06 UTC : members 21-40

12 UTC : members 41-60

18 UTC : members 61-80

I guess this was was chosen to try and maintain "forecast continuity". The naming convention for the products does not actually line up with the member number within the perturbation cycling. In other words, member NN" is not the same cycled "member NN" between each of the 4 cycles/day.

I suppose then it's possible through the breeding and selection process there could be relationships between the members (with a 24h lag though, not from cycle to cycle). This has become less likely now that they have implemented stochastic noise in the ensemble model integration.

Thanks! Great description.

[usedtobe]

everybody jumps in on my HPC post

obviously I was kidding, but you have to admit that's on he!! of a storm track to pick from out of all the variety of possibilities the models have shown

I was defending two friends and didn't realize it was a joke.

[Feb]

I was defending two friends and didn't realize it was a joke.

In all honesty i didn't take it as a joke either. I would not want to be a forecaster right now. This is where we weenies should step aside and let the pro's do their job.

[Riptide]

Sounds like "more of the same", with a non-cooperating NAO and la nina.

[Scarlet Pimpernel]

I was defending two friends and didn't realize it was a joke.

Not to pile on so much, but I didn't see it as a joke either. In the time I was at NCEP, I know those guys at the medium range desk have a tough job much of the time. Especially given how the guidance has been this winter! I never was on the "hot seat" in terms of having to put a forecast on the line like that, but I can still respect the years of expertise they bring to the table. Nobody is there to say "how can we hype up the snow weenies..mwaaa ha ha<insert Dr. Evil laugh>!!" They're professionals, making the best judgement that they can with the information available. Sometimes, it's really tough to sift through what that information is telling you. (Having done some forecasting years ago, I know well what that's like).

[Wonderdog]

You ask and you shall receive. First, you're basically correct....and from a scientific (and design) standpoint, every member should be an equally likely outcome.

However, the GEFS system does have one quirk relative to the other major operational centers that can lead to some correlative skill characteristics by "member number" (albeit with a 24 hour lag). The GEFS initial perturbations are applied using something called an ensemble transform (rescaled) method, which for all intents and purposes is a variant of breeding. For each cycle (00, 06, 12, 18), there is actually an 80 member ensemble of short (06h) forecasts. Those 80 forecast perturbations are then rescaled (shrunk) to represent initial condition perturbations (they are also reorthogonalized). From the 80 initial perturbations, only 20 are selected and integrated out as part of the EPS. The problem is that the selection process isn't random:

00 UTC : members 01-20

06 UTC : members 21-40

12 UTC : members 41-60

18 UTC : members 61-80

I guess this was was chosen to try and maintain "forecast continuity". The naming convention for the products does not actually line up with the member number within the perturbation cycling. In other words, member NN" is not the same cycled "member NN" between each of the 4 cycles/day.

I suppose then it's possible through the breeding and selection process there could be relationships between the members (with a 24h lag though, not from cycle to cycle). This has become less likely now that they have implemented stochastic noise in the ensemble model integration.

Thanks guys. I was wondering about the names and if they were the same every run. I guess that explains why each run at least this year, seems to indicate a different outcome on the three temperature maps (days 1-5; 6-10; and 11-15) that accompany each run.

[mitchnick]

I was defending two friends and didn't realize it was a joke.

meh to internet boards

I hate e-mails too

this world is going to he!! because of the lack of personal contact, and I'm not kidding about that

but we'll survive

[Midlo Snow Maker]

12z euro has a storm hr 144 in miss. some nice cad showing up in va.

[yoda]

12z euro has a storm hr 144 in miss. some nice cad showing up in va.

Talk to me Midlo...

[Midlo Snow Maker]

hr 156 look like some frozen stuff to start then 1 low heads up into eastern ky. and another broad area of low pressure on the south east coast looks miller b-ish

[yoda]

hr 156 look like some frozen stuff to start then 1 low heads up into eastern ky. and another broad area of low pressure on the south east coast looks miller b-ish

Sounds like a redeveloper... is the H strong to our north or weak?

[BethesdaWX]

I don't think latest the GEFS could get any better in the LR, wow

http://raleighwx.ame...alyNH_Loop.html

[Midlo Snow Maker]

weak low hardly any precip some light snow pa & ny........... ric and dc are warm +4 ~ +6

signal is there the details will take a few days atleast it's something to watch

[usedtobe]

weak low hardly any precip some light snow pa & ny........... ric and dc are warm +4 ~ +6

signal is there the details will take a few days atleast it's something to watch

It makes me worry a little about ice knowing how the euro warms up the 2m temps. Still it's so far out there in time, lots can change. It does support the HPC forecast idea.

[Srain]

WEATHER RECONNAISSANCE FLIGHTS

CARCAH, NATIONAL HURRICANE CENTER, MIAMI, FL.

0115 PM EST SUN 29 JANUARY 2012

SUBJECT: WINTER STORM PLAN OF THE DAY (WSPOD)

VALID 30/1100Z TO 31/1100Z JANUARY 2012

WSPOD NUMBER.....11-060

I. ATLANTIC REQUIREMENTS

1. NEGATIVE RECONNAISSANCE REQUIREMENTS.

2. OUTLOOK FOR SUCCEEDING DAY.....NEGATIVE.

II. PACIFIC REQUIREMENTS

1. FLIGHT ONE -- TEAL 72

A. P54/ DROP 8 (45.0N 166.0W)/ 31/0000Z

B. AFXXX 08WSC TRACK54

C. 30/1830Z

D. 17 DROPS AS PUBLISHED ON TRACK

E. AS HIGH AS POSSIBLE/ 31/0600Z

2. SUCCEEDING DAY OUTLOOK:

A. POSSIBLE TEAL C-130J MISSION FOR

P56/ DROP 9 (44.3N 151.0W)/ 01/0000Z.

B. POSSIBLE NOAA G-IV MISSION FOR

P56/ DROP 9 (44.3N 151.0W)/ 01/1200Z.

[Midlo Snow Maker]

what ever happens seems like a slow mover per this run that ridge out west goes all the way over the arctic ocean into russia 570dm into western canada

[Midlo Snow Maker]

hr 204 huge +pna nice -nao, big trough in the east

[BethesdaWX]

Arctic Attack at day 9 with a raging +PNA, -AO/-NAO. Right on que with the February 7th "end transition" period.

[usedtobe]

hr 204 huge +pna nice -nao, big trough in the east

But no southern stream and the trough is too far east to develer much more than cold air unless the miller b delivers. It's a very cold look. There hasn't been much consistency except for the love of a potent arctic connected PNA pattern.

[Midlo Snow Maker]

day 10 has -16c dc -20 philly -AO -NAO +PNA the PV is just south of hudson bay and looks like some hint of a split flow

[jviper]

Brrr

[usedtobe]

It is better and that's what's important. The strong -EPO look is really no longer there though the last 2 runs And the block is east of Iceland. It is serviceable though and we probably will get cold.

It's not a bad pattern going forward,certainly better than anything we've seen so far this year.