Jump to content
  • Member Statistics

    17,584
    Total Members
    7,904
    Most Online
    23Yankee
    Newest Member
    23Yankee
    Joined

Hurricane Irene Model and Forecast Discussion


Recommended Posts

  • Replies 1.3k
  • Created
  • Last Reply

NHC put E MA into a hurricane warning during Earl last year...which in my mind was possibly the most inexplicable hurricane warning ever issued there. I do not want to turn this into a lame NHC bash thread, but that was possibly the worst I have ever seen. Almost every single New England meteorologist was saying how even hurricane conditions were almost nil unless perhaps you were on ACK (nantucket) and even that was a long shot. Just a horrible piece of public awareness report...I know they have protocol for their cone of uncerntainty, but that was a perfect example of why its so flawed. We had news stations obsesssing over the hurricane warning while every met with a brain between their ears was saying this was absolutely no big deal and just some heavy rain with 30 knot gusts...nevermind sustained winds.

kush and I argued about this a month ago. I strongly believe the cone and warning criteria should be Bayesian based on the model clustering. Mark DeMaria is working on the problem out at CIRA right now.

Link to comment
Share on other sites

Again note the HUGE differences going on out west with the second kicker. I'm growing more and more convinced THAT is the feature we should be following. God this is just like winter where stuff all comes down to Pacific shortwaves where we have literally have no sampling...

Regarding radiosonde obs, yes no sampling, but there are millions of satellite obs from over the oceans ingested for each cycle. Having said that, sonde data is higher quality and weighted more by the models because of that, but the Pacific shortwaves are sampled to some extent.

Link to comment
Share on other sites

Regarding radiosonde obs, yes no sampling, but there are millions of satellite obs from over the oceans ingested for each cycle. Having said that, sonde data is higher quality and weighted more by the models because of that, but the Pacific shortwaves are sampled to some extent.

Hopefully dtk can jump in here, but he and I had some discussions in winter about this. The rawindsonde network sampling isn't that big of a deal EXCEPT in high impact situations where the eventual outcome is sensitive to a few meters here or there in the initialization of the s/w.

Link to comment
Share on other sites

Again note the HUGE differences going on out west with the second kicker. I'm growing more and more convinced THAT is the feature we should be following. God this is just like winter where stuff all comes down to Pacific shortwaves where we have literally have no sampling...

28jxwsy.gif

It actually looks like the ECMWF looks more like the 00Z GFS than the 06Z GFS with regard to the strength of that wave dropping into the north central US.

Link to comment
Share on other sites

Nice HPC disco discussing the influence of the northern stream Canadian wave currently amplifying over the GL and the possible influence on Irene.

PRELIMINARY EXTENDED FORECAST DISCUSSION

NWS HYDROMETEOROLOGICAL PREDICTION CENTER CAMP SPRINGS MD

917 AM EDT WED AUG 24 2011

VALID 12Z SUN AUG 28 2011 - 12Z WED AUG 31 2011

...DANGEROUS HURRICANE IRENE TO AFFECT THE NC OUTER BANKS INTO THE

NORTHEAST THIS WEEKEND...

DUE IN PART TO MORE FREQUENT SAMPLING OF THE ENVIRONMENTAL

CONDITIONS SURROUNDING HURRICANE IRENE BY MULTIPLE AIRPLANE

PLATFORMS ALONG WITH CONTINUATION OF SUPPLEMENTAL UPPER AIR RAOBS

ACROSS PARTS OF THE EASTERN CONUS EVERY 6 HOURS...THE MODEL

SOLUTION SPREAD FOR THE TRACK OF IRENE HAS BECOME SMALLER AND

SMALLER...AND IN FACT COULD BE CONSIDERED EXCELLENT FOR THE LONGER

RANGES. ONE OF THE LARGEST REMAINING UNCERTAINTIES IN THE MEDIUM

RANGE PERIOD HOWEVER IS THE EXACT SPEED AND STRENGTH OF A STRONG

SHORTWAVE TROUGH CURRENTLY ENTERING WESTERN CANADA...WHICH IS

EXPECTED TO AMPLIFY OVER ONTARIO/QUEBEC BY DAY 3/SAT AND PICK UP

IRENE. THE 00Z GFS/ECMWF HAVE NEARLY CONVERGED WITH THE DETAILS OF

THIS TROUGH...WITH THE GFS ONLY SLIGHTLY FASTER AND SUPPORTED BY

THE CANADIAN...THUS ACCELERATING IRENE NORTHWARD INTO THE

NORTHEAST/NEW ENGLAND MORE QUICKLY THAN THE ECMWF.

HOWEVER...RUN-TO-RUN VARIABILITY WITH THE MORE UNCERTAIN HIGHER

LATITUDE FLOW IS TO BE EXPECTED...WITH RECENT GFS/ECMWF 4-CYCLE

LAGGED AVERAGE FORECASTS SHOWING 500 MB HEIGHT VARIABILITIES

AVERAGING 50 TO 70 DECAMETERS ACROSS THE UPPER GREAT

LAKES...ENOUGH TO SUPPORT DIFFERING INFLUENCES ON IRENE. BUT FOR

NOW...THE GFS/ECMWF APPEAR TO BE IN RELATIVELY GOOD AGREEMENT WITH

IRENE TO USE NEARLY EQUALLY FOR THE PRELIMINARY PRESSURES/FRONTS

FOR DAYS 3-7. THIS APPROACH WORKS WELL FOR THE WESTERN AND CENTRAL

CONUS AS WELL...AS THE STRONG FLOW ACROSS THE NORTHERN PACIFIC IS

EXPECTED TO SUPPORT CONTINUED STRENGTH AND TIMING DIFFERENCES OF

INDIVIDUAL SHORTWAVE TROUGHS WHICH ARE BEST RESOLVED WITH AN

INTERMEDIATE APPROACH. ONE DIFFERENCE SEEN IN THE LATEST GUIDANCE

IS FOR GREATER AMPLIFICATION OF A SIGNIFICANT SHORTWAVE TROUGH

ACROSS THE PACIFIC NORTHWEST DAYS 6/7...WITH THE ECMWF FASTER IN

EJECTING THE LEADING EDGE OF THE TROUGH OUT OF THE NORTHWEST

COMPARED TO THE GFS...WHICH THEN LEADS TO GROWING DIFFERENCES

DOWNSTREAM. WITH THE 00Z CANADIAN AND THE GLOBAL ENSEMBLE MEANS

PROVIDING GOOD SUPPORT FOR THE GFS SOLUTION...WILL LIKELY ABANDON

THE ECMWF ALTOGETHER STARTING DAY 6...AND INSTEAD BLEND THE 00Z

GFS/GEFS MEAN SOLUTIONS FOR DAYS 6/7.

JAMES/SCHICHTEL

Link to comment
Share on other sites

I'd be interested in hearing what you mean, or maybe you could link to a paper.

Right now the cones and warning criteria are "dumb" (I mean that in a statistical sense). They use the global mean errors to compute the width of the cone. But obviously, some storms have more uncertain forecasts than others. Earl was relatively easy, while Irene has proven to be relatively hard. Using the model spread to calculate the cone gives a "truer" idea of what the uncertainty is. However, as kush pointed out, there are psychological aspects to varying the cone that I don't understand well.

Link to comment
Share on other sites

Why have the actual "hurricane models" been so different this whole storm then the regular models?

I am not sure how "scientific" this explanation is (or if stats would back it up), but my experience is these large storms like Irene tend to be better handled by the global models, possibly because there is more dynamic interaction between the storm itself and the atmosphere. Conversely, smaller hurricanes tend to be handled better by the hurricane models than the global models, probably because the global models cannot grasp the tight, deep system.

Link to comment
Share on other sites

That's not the ensemble mean. Is it just another version of the GFS? Never came across this model until Irene.

It's the exact same model physics as the operational GFS, it just uses a hybrid ensemble Kalman filter initialization scheme versus standard 3DVAR for the operational. It's been performing well this year in the WPAC - better than the ops. dtk can provide more info if he's around today.

Link to comment
Share on other sites

Yes. It's the initialization scheme, versus the GFS ops using 3DVAR

Thanks for the explanation. Usually when I see "KF" with a model I think "Kain-Fritsch" convective scheme, and was wondering how that could be making that much of a difference. :arrowhead:

Link to comment
Share on other sites

Right now the cones and warning criteria are "dumb" (I mean that in a statistical sense). They use the global mean errors to compute the width of the cone. But obviously, some storms have more uncertain forecasts than others. Earl was relatively easy, while Irene has proven to be relatively hard. Using the model spread to calculate the cone gives a "truer" idea of what the uncertainty is. However, as kush pointed out, there are psychological aspects to varying the cone that I don't understand well.

So you mean using the spread as the prior in subsequent periods to incorporate uncertainty? Interesting.

Link to comment
Share on other sites

It's the exact same model physics as the operational GFS, it just uses a hybrid ensemble Kalman filter initialization scheme versus standard 3DVAR for the operational. It's been performing well this year in the WPAC - better than the ops. dtk can provide more info if he's around today.

Bleah! I didn't want to see "Kalman Filter" here - I've spent the past week trying to get a Kalman filter for sensor contact/track correlation to work in a naval warfare model, to the point where I'm dreaming about it.

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...