Jump to content
  • Member Statistics

    15,470
    Total Members
    7,904
    Most Online
    Greenmachine1123
    Newest Member
    Greenmachine1123
    Joined
HoarfrostHubb

Spring 2019 New England Banter and Disco

Recommended Posts

Just now, OceanStWx said:

Overall it's a marginal improvement over the GFS, and we shouldn't expect to see drastic changes from the old core. 

I've been thinking about this too...is it possible that increasing resolution size and layers could actually reduce the accuracy of models (unless that model is specifically designed for mesoscale?) I feel like if the resolution becomes too granular it may start picking up on certain "things" which could totally screw up the projection? Unless of course I guess this was adjusted for in the incredibly complex codes...but overall there have been these significant changes to the GFS with regards to resolution with not much improvement (or what would be thought) over the last several years. 

Share this post


Link to post
Share on other sites

Although this is probably the first time I've never not seen the GFS show massive convective blobs past like 200 hours. 

Share this post


Link to post
Share on other sites
10 minutes ago, weatherwiz said:

I've been thinking about this too...is it possible that increasing resolution size and layers could actually reduce the accuracy of models (unless that model is specifically designed for mesoscale?) I feel like if the resolution becomes too granular it may start picking up on certain "things" which could totally screw up the projection? Unless of course I guess this was adjusted for in the incredibly complex codes...but overall there have been these significant changes to the GFS with regards to resolution with not much improvement (or what would be thought) over the last several years. 

Well the resolution didn't change this time around.

But overall that's why they run these things in parallel, so they can test that they aren't harming any previous good a model did. This is a "do no harm" upgrade, so any degradation is basically noise. There will be plenty of hot takes about how bad the model is when it has its first big bust, but those will be mostly overblown beyond the normal the GFS just isn't as good a global model as the Euro and Ukie. 

Share this post


Link to post
Share on other sites
Just now, OceanStWx said:

Well the resolution didn't change this time around.

But overall that's why they run these things in parallel, so they can test that they aren't harming any previous good a model did. This is a "do no harm" upgrade, so any degradation is basically noise. There will be plenty of hot takes about how bad the model is when it has its first big bust, but those will be mostly overblown beyond the normal the GFS just isn't as good a global model as the Euro and Ukie. 

ahhh I thought the resolution and amount of layers broken into was increased. 

All in all though this has been in testing for quite some time and I'm sure it would not become operational if there were significant issues. 

Share this post


Link to post
Share on other sites
Just now, OceanStWx said:

Well the resolution didn't change this time around.

But overall that's why they run these things in parallel, so they can test that they aren't harming any previous good a model did. This is a "do no harm" upgrade, so any degradation is basically noise. There will be plenty of hot takes about how bad the model is when it has its first big bust, but those will be mostly overblown beyond the normal the GFS just isn't as good a global model as the Euro and Ukie. 

Don't the new equations enhance the resolution to a certain degree though despite it still having a resolution of 13km in the horiztonal and 64 in vertical?

Share this post


Link to post
Share on other sites
Just now, weatherwiz said:

ahhh I thought the resolution and amount of layers broken into was increased. 

All in all though this has been in testing for quite some time and I'm sure it would not become operational if there were significant issues. 

I read that in several cases it outperformed the Euro for tropical systems in regards to model forecast error so I'm expecting good things with it. Here is one article I read that highlights a lot of new features: https://www.weather.gov/media/notification/scn19-40gfs_v15_1.pdf

Share this post


Link to post
Share on other sites
1 minute ago, It's Always Sunny said:

Don't the new equations enhance the resolution to a certain degree though despite it still having a resolution of 13km in the horiztonal and 64 in vertical?

Certain parameters may improve (like the satellite microphysics I mentioned) but like you say the grid resolution won't. This upgrade will make it easier to change resolution down the road though, which is one of the major reasons for the upgrade.

Share this post


Link to post
Share on other sites

Any particular reason the,fv3 maintains high resolution beyond D10? Worth the delay and server load?

Share this post


Link to post
Share on other sites
5 minutes ago, Dr. Dews said:

Any particular reason the,fv3 maintains high resolution beyond D10? Worth the delay and server load?

More gridpoints for wizzy to point and click as the error approaches the limit of infinity.

  • Haha 2

Share this post


Link to post
Share on other sites
25 minutes ago, It's Always Sunny said:

I read that in several cases it outperformed the Euro for tropical systems in regards to model forecast error so I'm expecting good things with it. Here is one article I read that highlights a lot of new features: https://www.weather.gov/media/notification/scn19-40gfs_v15_1.pdf

Very interesting! Thanks for sharing. 

2 minutes ago, dendrite said:

More gridpoints for wizzy to point and click as the error approaches the limit of infinity.

:weenie: 

Share this post


Link to post
Share on other sites

I have the PDF demo for it but the site says it exceeds the file size limit. Pretty comprehensive comparison. 

Share this post


Link to post
Share on other sites
11 hours ago, dendrite said:

More gridpoints for wizzy to point and click as the error approaches the limit of infinity.

May I rant, about the non-scientific nature?

Share this post


Link to post
Share on other sites

Right out of the gate, our 12 hr differencing graphics were colder in every panel yesterday...lol. Either the cold bias wasn't completely removed, or the old gfs had that warm bias it can carry. Perhaps a bit of both. 

Share this post


Link to post
Share on other sites

Hm .. the former GFS had multiple biases, each one could skew reality in their own right and at times mimick the other bias' ... at other times, offset.  Really frustrating that way... 

For example, the model's progressive bias - it ablates the tops of ridges down too quickly in the mid range.  This has feed backs that cause new errors... for one, ends up increasing confluence over eastern Canada ( just one example in many ...), which then we have overly strong BD/west moving CAD signals...  That is a cold bias result, in a model with a warm boundary layer problem.   Zoink :wacko2:

There's other areas where is seems to contradictory bias its self...  Maybe that fooled the testers/QC evaluators ( ha ha...kidding here)  because between heat wave boundary layers of 114 F at Nashua NH, while given least excuse imaginable to snow in D.C. from a BD cold air mass in July ... you end up with the right temperature for NYC.   Yea ...see?  good model.  

Share this post


Link to post
Share on other sites
6 minutes ago, Typhoon Tip said:

Hm .. the former GFS had multiple biases, each one could skew reality in their own right and at times mimick the other bias' ... at other times, offset.  Really frustrating that way... 

For example, the model's progressive bias - it ablates the tops of ridges down too quickly in the mid range.  This has feed backs that cause new errors... for one, ends up increasing confluence over eastern Canada ( just one example in many ...), which then we have overly strong BD/west moving CAD signals...  That is a cold bias result, in a model with a warm boundary layer problem.   Zoink :wacko2:

There's other areas where is seems to contradictory bias its self...  Maybe that fooled the testers/QC evaluators ( ha ha...kidding here)  because between heat wave boundary layers of 114 F at Nashua NH, while given least excuse imaginable to snow in D.C. from a BD cold air mass in July ... you end up with the right temperature for NYC.   Yea ...see?  good model.  

I hope someone is eventually able to get the issues fixed too regarding the NAM being too moist with profiles during the overnight. 

Share this post


Link to post
Share on other sites

I see his point though ... 

I had this conversation with Ekster years ago...when the ETA was just becoming the NAM, how the model was kind of like a 'victim of its own success'.  

The problem with immense computing power operating on very small grid spacing ...means you're starting to tap into the uncertainty principle - as a metaphor.  I don't mean really down at the quantum scales.. .but.. .in some ways, something similar happens.  Just apply electrons to a system that 'predicts' based upon very finite inputs for finite scales, you end up with overly emphasized forces and factors that "giga" motion the system into unwanted results - in effect, you open Pandora's box of fractals.   .

Models can't ultimately predict the future? 

That's not really what's happening when the models are fired off...  They are predicting likely outcomes, with ever decreasing probability for success in doing so at that, for every quantum instant of time that elapses further and further into the future. But the actual reality of the future cannot really be ascertained without actually being in the future, because of mercurial nature of ( almost ) unavoidable chaos.  

By the time we get to day fives, we're ...I dunno somewhere in the 40th to 60th percentile for success, and it's dependent upon the 'stability' of the pattern at hand there. 

Day ten?  Forget it... 10 maybe 20% tops.. Which means, by those deeper range time spans the unpredictable, unknowable future circumstances that emerge along the way ( chaos ) have corrupted the futility of models down to guess work, all but entirely.

There is a theoretical limit to that success rate. Models can max out.  We're not there yet.  But techniques, such as ( maybe ) the Euro 4-d normilzation schemes, which are remarkably successful at picking and choosing those spontaneously emergent distractions that need to be  'canceled out', can be applied to models with NAM-like finite grids.  Who knows..just spit-ballin' there.  But there's room to improve ... and all those improvements combined, we'll never get 100% accurate at some theoretical limit, because ultimately ... the uncertainty of chaos cannot be preordained. 

The only way to do so ... as hinted by that parenthetical 'almost' above ... is to control the future.  Science fiction ...for now.  But, if there can be conjured technology that governs the quantum momentum state of every particle that embodies the fluid medium of the atmosphere, sufficiently that it suppresses "butterflies" .. then you don't have to predict the weather: the solution is, push this button if you want a sunny day. But you know what's funny ...?  In a philosophical sense, even in such a fantasy futuristic world, there is uncertainty in a system that 100% capable of modulating the weather.   Because there's no guarantee that the operator won't be influenced by either a foreign agent, or lapse into some sort of psychosis that entices him/her to push typhoon buttons. 

God have mercy on those souls in that realm of existence.   

  • Like 2

Share this post


Link to post
Share on other sites
5 minutes ago, Typhoon Tip said:

I see his point though ... 

I had this conversation with Ekster years ago...when the ETA was just becoming the NAM, how the model was kind of like a 'victim of its own success'.  

The problem with immense computing power operating on very small grid spacing ...means you're starting to tap into the uncertainty principle - as a metaphor.  I don't mean really down at the quantum scales.. .but.. .in some ways, something similar happens.  Just apply electrons to a system that 'predicts' based upon very finite inputs for finite scales, you end up with overly emphasized forces and factors that "giga" motion the system into unwanted results - in effect, you open Pandora's box of fractals.   .

Models can't ultimately predict the future? 

That's not really what's happening when the models are fired off...  They are predicting likely outcomes, with ever decreasing probability for success in doing so at that, for every quantum instant of time that elapses further and further into the future. But the actual reality of the future cannot really be ascertained without actually being in the future, because of mercurial nature of ( almost ) unavoidable chaos.  

By the time we get to day fives, we're ...I dunno somewhere in the 40th to 60th percentile for success, and it's dependent upon the 'stability' of the pattern at hand there. 

Day ten?  Forget it... 10 maybe 20% tops.. Which means, by those deeper range time spans the unpredictable, unknowable future circumstances that emerge along the way ( chaos ) have corrupted the futility of models down to guess work, all but entirely.

There is a theoretical limit to that success rate. Models can max out.  We're not there yet.  But techniques, such as ( maybe ) the Euro 4-d normilzation schemes, which are remarkably successful at picking and choosing those spontaneously emergent distractions that need to be  'canceled out', can be applied to models with NAM-like finite grids.  Who knows..just spit-ballin' there.  But there's room to improve ... and all those improvements combined, we'll never get 100% accurate at some theoretical limit, because ultimately ... the uncertainty of chaos cannot be preordained. 

The only way to do so ... as hinted by that parenthetical 'almost' above ... is to control the future.  Science fiction ...for now.  But, there can be conjured technology that governs the quantum momentum state of every particular that comprises the fluid medium of the atmosphere, sufficiently that it suppresses "butterflies" .. then you don't have to predict the weather: the solution is, push this button if you want a sunny day. 

God have mercy on those souls in that realm of existence.   

This is a terrific, terrific statement IMO...and extremely underrated. Anyone who views computer forecasts models should have to read this disclaimer prior to proceeding...seriously. This is what separates model huggers from the rest. You need to use experience and knowledge to determine whether the solution has merit...let's say model X keeps showing a snowstorm...but there is a good deal of spread/uncertainty...by analyzing large scale features you *should* be able to gauge that actual probability of that solution verifying and determine whether to toss or if it has merit. 

Share this post


Link to post
Share on other sites

Looks like a wet Father's Day incoming. Did some surf casting last night...one schoolie in our group but pretty meh overall. It was chilly earlier in the evening and then the wind switched more W-SW and felt like someone turned on the hair dryer. Love that effect.

Share this post


Link to post
Share on other sites

Anyone along and N of this line is polishing a turd to call this season summer ( so far .. )

 

image.png.0fdae7cf803a16bb05d701687baf9d08.png

What's happened so far is winter with June sun shining through it ...

  • Like 1

Share this post


Link to post
Share on other sites
57 minutes ago, Typhoon Tip said:

Anyone along and N of this line is polishing a turd to call this season summer ( so far .. )

image.png.0fdae7cf803a16bb05d701687baf9d08.png

Right from the opening bell Memorial day weekend I'd have called this stretch just about as good as it gets for early summer.  Not sure what folks are looking for...

  • Like 1

Share this post


Link to post
Share on other sites
On 6/15/2019 at 6:01 PM, yoda said:

Does Dr. Dews change his avatar every day?

Same schedule as his panties.

  • Haha 2

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Recently Browsing   0 members

    No registered users viewing this page.

×