Model Mezzanine, the 2nd installment

[OceanStWx]

Here's the GFS forecast initialized 00z 2/16 for PWM

 

DAY / HOUR   16/00  16/06 16/12
------------------- ------   ------   ------ 
TEMPS
SFC (2 M) (F)    21       31      37

 

Actual                21       20     27

[CoastalWx]

It is what it is, an inferior model to the Euro. It's not all of a sudden going to start consistently banging out better verification.

Its QPF fields can make sense. But it's really an uneasy feeling when I have to toss a major piece of guidance and go with gut on thermals. It's unnerving because I'm always asking myself what if it is right. Most of the time, it is not.

[RUNNAWAYICEBERG]

GEFS are pretty nice.

 

the mean is pretty far se of 12z gefs. i guess it followed the "trend" of the op to a certain extent. 

[OSUmetstud]

Here's the GFS forecast initialized 00z 2/16 for PWM

 

DAY / HOUR   16/00  16/06 16/12

------------------- ------   ------   ------ 

TEMPS

SFC (2 M) (F)    21       31      37

 

Actual                21       20     27

It doesn't seem to matter where u are.  warm everywhere in the low levels.  I have icebergs out my backdoor and it still wants to warm me up into the mid 30s with e or ne winds...yah right.  

 

(Iceberg pics coming soon...I missed them yesterday.)

[powderfreak]

It is what it is, an inferior model to the Euro. It's not all of a sudden going to start consistently banging out better verification.

The GFS seems much more prone to wild swings too from run to run. The EURO almost seems like it has some algorithm that doesn't allow for huge changes run to run, but a more gradual trend. Like this last storm we saw the GFS go from a weak POS low tracking over Boston on an 18z run turn into a 985mb bomb over SYR at 00z, all while the EURO actually very slowly moved east from a low over ROC to a low over ALB in the days leading up.

[OceanStWx]

Not a bad approach. I just think if you know your "base audience" up here (every skier and snowboarder is monitoring every keystroke of the NWS), and yes I think those with outdoor recreation interests are their biggest group, you are going to make damn sure it will snow before getting folks hopes up. It's almost like a Blizz/DIT devil's advocate scenario...if you say it won't snow and it does you still "win" for a huge portion of the population up here. From hotels, lodging, shops, anyone related to tourism, 75% of the public skis/snowboards/snowmobiles, everyone would be much more pleased with a positive snow bust than the opposite. Probably one of the rare areas where a social media poll might actually show that, haha.

If you say it's going to snow and it doesn't, I think you're more likely to hear about it. This isn't Boston where if you mess up the evening commute forecast you get castrated, haha.

 

Except the major issue for us is that if we say it won't snow and then there is a 20 car pile up with fatalities on 89, we have to answer to someone about why it wasn't in the forecast. And not just in the near term, they want to know if it was highlighted 4, 5, 6 days in advance.

[OceanStWx]

Its QPF fields can make sense. But it's really an uneasy feeling when I have to toss a major piece of guidance and go with gut on thermals. It's unnerving because I'm always asking myself what if it is right. Most of the time, it is not.

 

I know. In fact I just created an edit area for CAD based on observed temps the other day, just because I know that models will consistently warm up near coastal areas too much in the middle ranges. Once we get to near term we have enough mesoscale guidance that picks up on it, but days 3-5 when we can add details that are useful, our 2 m temp guidance is putrid (remember even the "better" Euro is 80 km resolution for us).

[RUNNAWAYICEBERG]

for visual purposes for peeps who cant find maps:

 

18z

 

 

12z 

 

[RUNNAWAYICEBERG]

There are some really sweet members with BM tracks though. I'll lean towards those haha. 

[Damage In Tolland]

Pretty big jump east

[CoastalWx]

I know. In fact I just created an edit area for CAD based on observed temps the other day, just because I know that models will consistently warm up near coastal areas too much in the middle ranges. Once we get to near term we have enough mesoscale guidance that picks up on it, but days 3-5 when we can add details that are useful, our 2 m temp guidance is putrid (remember even the "better" Euro is 80 km resolution for us).

 

Definitely. Of course on the East Coast when you have terminals naturally close to the baroclinic zone, it means a lot. I have constantly tried to insert my thoughts on GFS or incorporating GFS output for snow/ice etc. It's not an easy task. At the same time, I have really liked the QPF output or at least the idea of it. So, toss the thermals, but ride the QPF? LOL. Sometimes that is what you do. 

[OSUmetstud]

The ice risk is there with the huge Bermuda-Azores high pumping warm air...partly the same reason that Ottawa got one of it's biggest snowstorms in history yesterday.  

[ORH_wxman]

I know. In fact I just created an edit area for CAD based on observed temps the other day, just because I know that models will consistently warm up near coastal areas too much in the middle ranges. Once we get to near term we have enough mesoscale guidance that picks up on it, but days 3-5 when we can add details that are useful, our 2 m temp guidance is putrid (remember even the "better" Euro is 80 km resolution for us).

 

 

Even the higher res Euro doesn't always do the greatest with CAD...it shows it, but is often too far north or erodes it too quickly. Way better than the GFS though.

 

It just seems to be a symptom of every model though. They can't hold the CAD long enough and often not far enough south. RGEM is probably one of the best once we get close. NAM used to be amazing, but I have seen some bad performances by it ever since they upgraded it in 2012. Not sure why.

 

I remember how good the NAM did in the 2008 ice storm.

[HIPPYVALLEY]

Pretty big jump east

Congrats James?

[CoastalWx]

We ride BTV WRF thermals.

[bobbutts]

It seems you could make fast improvement if you had the computing power to run basically unlimited perturbations of previous model runs against actual results and some fancy "AI" to analyze the results.. I gather it's cost prohibitive to do that.

[40/70 Benchmark]

The thing with "Tendencies" is that why should we use them?

 

Last year everything went right before 3 weeks and then it changed in late Feb/Mar....in 2011 we had epic January and early February, then it changed....in 2006-2007 we had a deplorable first half and then it changed. In 2012-2013 we had an awful first half (on the coast anyway, not interior) and then it changed.

Don't use them.

I never said anyone should..

Sometimes I choose to ....its always a risk, as in any call at day 6.

[OSUmetstud]

Even the higher res Euro doesn't always do the greatest with CAD...it shows it, but is often too far north or erodes it too quickly. Way better than the GFS though.

 

It just seems to be a symptom of every model though. They can't hold the CAD long enough and often not far enough south. RGEM is probably one of the best once we get close. NAM used to be amazing, but I have seen some bad performances by it ever since they upgraded it in 2012. Not sure why.

 

I remember how good the NAM did in the 2008 ice storm.

try the higher res one.  It's run at 2.5 km.  It's on wxbell.  Like any meso gets qpf happy but good for CAD.  

[OceanStWx]

The GFS seems much more prone to wild swings too from run to run. The EURO almost seems like it has some algorithm that doesn't allow for huge changes run to run, but a more gradual trend. Like this last storm we saw the GFS go from a weak POS low tracking over Boston on an 18z run turn into a 985mb bomb over SYR at 00z, all while the EURO actually very slowly moved east from a low over ROC to a low over ALB in the days leading up.

 

This is where I can be slightly out of my depth at times, but I believe there may actual be a real reason for that. I'm pretty sure I remember hearing that part of the Euro initialization scheme is to go much farther back in time and begin the model, adjusting for what we know has already happened. That was at 00 hour it's got a better handle on past weather, and can use that knowledge to forecast that future.

 

Say the 00z GFS is initialized. Well it actually uses information from the previous model run for 00 hour, and it doesn't start "cold" at that time either, but say at 12z and runs through 00 hour. The Euro on the other hand may actually start at 12z the day before and run for 24 or 48 hours before getting to 00 hour. Something like that.

 

A "cold" start would probably be more likely to swing wildly, than something being "guided" through start time.

[CoastalWx]

This is where I can be slightly out of my depth at times, but I believe there may actual be a real reason for that. I'm pretty sure I remember hearing that part of the Euro initialization scheme is to go much farther back in time and begin the model, adjusting for what we know has already happened. That was at 00 hour it's got a better handle on past weather, and can use that knowledge to forecast that future.

 

Say the 00z GFS is initialized. Well it actually uses information from the previous model run for 00 hour, and it doesn't start "cold" at that time either, but say at 12z and runs through 00 hour. The Euro on the other hand may actually start at 12z the day before and run for 24 or 48 hours before getting to 00 hour. Something like that.

 

A "cold" start would probably be more likely to swing wildly, than something being "guided" through start time.

 

I think that's part of the 4-D VAR. I know it also includes real time obs after 12z as the model assimilates too, unlike GFS.

[OceanStWx]

I think that's part of the 4-D VAR. I know it also includes real time obs after 12z as the model assimilates too, unlike GFS.

 

Part of our computing resource shortage. Can't 4-DVAR the GFS when you run the SREF 4 times a day, etc. 

[40/70 Benchmark]

How much snow did Ottawa receive?

[Morch Madness]

GEFS look pretty cold and active after the storm as well

[Hoth]

It seems you could make fast improvement if you had the computing power to run basically unlimited perturbations of previous model runs against actual results and some fancy "AI" to analyze the results.. I gather it's cost prohibitive to do that.

Give it ten years.

[RUNNAWAYICEBERG]

How much snow did Ottawa receive?

 

51.2 cm or in our world, 20".

[ncforecaster89]

Looking at the NESIS site, it's pretty obvious why most major storms occur on a weekend...it's because they all affected the broader northeast region over 3 or more days. So you are already starting with odds of at least one weekend day falling within that period just based on the math of it.

Excellent observation, JC! I will note that I did notice that the peak effects of the majority of those events occured on a weekend...with some notable exceptions like the blizzards of 1978 and 2015. This is likely another reason why so many remember those storms occurring on a weekend...leading to the presumption of the "weekend rule."

[CoastalWx]

51CM of snow in Ottawa in less than 15hrs.

[codfishsnowman]

The GFS is such horsesh*t with the thermals. That high nosing in and it's barely cold from 1000-850 even in the Berks. Yeah, ok.  

if there are such issues with the thermal profiles why doesn't the government fix it??

[CoastalWx]

Excellent observation, JC! I will note that I did notice that the peak effects of the majority of those events occured on a weekend...with some notable exceptions like the blizzards of 1978 and 2015. This is likely another reason why so many remember those storms occurring on a weekend...leading to the presumption of the "weekend rule."

It's voodoo and obvious voodoo. Weather will do what it wants to do, regardless of aerosols.

[Ginx snewx]

Yeah that deform in ROC to Ottawa area put down serious paste