December Banter Thread 2

[H2O]

I'm gonna live my life normally regardless of what some old made up Greek superhero storm does

[Ian]

Can anyone name a few storms that changed markedly on models once the disturbance was in the us data network? Or rather can we be sure that's what set the course?  "Once this thing comes ashore we'll know a lot more" seems to be one of those things thrown about just because. 

[PhineasC]

Time to predict what the 00Z models will predict.

 

NAM will look much better (OK storm), but still be sort of silly and hard to believe.

GFS will come back towards the 12Z solution, but only half-way, leaving weenies feeling unsatisfied and concerned.

Euro will improve slightly, but horrible 1:30 AM analysis will leave everyone confused. 

ORH_wxman and/or CoastalWX will make an appearance and mention something about frontogenesis in DC.

 

Bonus:

 

GGEM will show a rainstorm.

UKMET will have a weak low heading OTS off the Florida coast.

[Bob Chill]

Nice, thanks. I'm going with #15 on snow probably.

 

Romo'd!

[Bob Chill]

Can anyone name a few storms that changed markedly on models once the disturbance was in the us data network? Or rather can we be sure that's what set the course?  "Once this thing comes ashore we'll know a lot more" seems to be one of those things thrown about just because. 

 

Not really. That weenie concept typically applies to storms that aren't working out at longer leads. Based on the last 3 years I can't think of one that made a damn bit of difference in the right direction. 

[Fozz]

Speaking of Nemo

 

nemo-forecasts.gif

 

[WxUSAF]

Can anyone name a few storms that changed markedly on models once the disturbance was in the us data network? Or rather can we be sure that's what set the course?  "Once this thing comes ashore we'll know a lot more" seems to be one of those things thrown about just because. 

I was just thinking about that as well.  I'm not sure I can remember a specific example.  It "makes sense" as an idea because the UA sampling is so much denser once a s/w comes onshore.  But there is data from over the Pac that is assimilated into the models.  Given how all the 12z models did seem to jump all together, I wondered if that was what happened this morning before I checked to see the s/w was still offshore. 

[mattie g]

Saints are in serious trouble in that one.

It's going to be a much better game than it would have been in Dallas if they had won last night, that's for sure. If the Eagles' D can do what they've been doing for much of the second half of the season (barring the Vikings game and some of last night), then it should be a gat game.

[mappy]

I'm gonna live my life normally regardless of what some old made up Greek superhero storm does

What?! That's not how it works. You must watch every model run, extrapolate it them before hrs come out, sing and dance for snow or having a mattieG type meltdown when it shows nothing.

[packbacker]

Not really. That weenie concept typically applies to storms that aren't working out at longer leads. Based on the last 3 years I can't think of one that made a damn bit of difference in the right direction.

Boxing Day 2010, I know that's a bad word on this forum but all models had the storm 3-5 days out, lost it, then got it back 36 hours out.

[Ian]

Not really. That weenie concept typically applies to storms that aren't working out at longer leads. Based on the last 3 years I can't think of one that made a damn bit of difference in the right direction.

You see it mentioned by gov and other mets too though. Intuitively you'd think more data is good but I can't remember any storm myself where as soon as it hit the west coast models changed drastically and locked in. In a situation like this maybe you could come to that conclusion as we've more or less been in model spray mode but is it better data or just closing in.

[mattie g]

What?! That's not how it works. You must watch every model run, extrapolate it them before hrs come out, sing and dance for snow or having a mattieG type meltdown when it shows nothing.

Burn!

My excuse is that my kid left us with negligible sleep and I couldn't hack it. I'm much more level-headed these days. I'm not invested at all in this one.

[Ian]

I was just thinking about that as well. I'm not sure I can remember a specific example. It "makes sense" as an idea because the UA sampling is so much denser once a s/w comes onshore. But there is data from over the Pac that is assimilated into the models. Given how all the 12z models did seem to jump all together, I wondered if that was what happened this morning before I checked to see the s/w was still offshore.

Satellites provide tons of data. In reality the models do quite well overall in areas that don't have the us observation capabilities. Obviously the ground and balloon truth is better than not having it.

[Ian]

Boxing Day 2010, I know that's a bad word on this forum but all models had the storm 3-5 days out, lost it, then got it back 36 hours out.

Was that because it came into the west coast? Were the models that were wrong the day before better than the ones that showed us getting screwed in days prior?

[H2O]

Can anyone name a few storms that changed markedly on models once the disturbance was in the us data network? Or rather can we be sure that's what set the course?  "Once this thing comes ashore we'll know a lot more" seems to be one of those things thrown about just because.

Jan 2000 comes to mind. Was supposed to be a meh storm and it flipped to a big storm inside of 24 hours or so

[WxUSAF]

Satellites provide tons of data. In reality the models do quite well overall in areas that don't have the us observation capabilities. Obviously the ground and balloon truth is better than not having it.

Indeed. If I remember I'll try to do a literature search on this idea tomorrow when I get back to a network with institutional subscriptions. 

[Bob Chill]

You see it mentioned by gov and other mets too though. Intuitively you'd think more data is good but I can't remember any storm myself where as soon as it hit the west coast models changed drastically and locked in. In a situation like this maybe you could come to that conclusion as we've more or less been in model spray mode but is it better data or just closing in.

 

I'm sure there's merit to it but down stream is already modeled at shorter leads with well samples UA analysis.  It's not like a vort was prog'd to hit bc but ended up at san fran. 

 

Sometimes they run recon when downstream is supportive of a high impact event. I don't think it suddenly changes anything. Maybe dials in finer details in strength and track. 

 

I might be totally out to lunch on the subject but my personal obs don't really give me confidence that all will be right with the world the second a sw hits the coast. 

[H2O]

What?! That's not how it works. You must watch every model run, extrapolate it them before hrs come out, sing and dance for snow or having a mattieG type meltdown when it shows nothing.

You're right. I'll go look at the 84 hour NAM now and lose my shiit no matter what it shows.

[Ian]

Indeed. If I remember I'll try to do a literature search on this idea tomorrow when I get back to a network with institutional subscriptions. 

After Haiyan I tried to track down the feasibility and usefulness of resuming some sort of recon (drone based).  I ended up talking to one of the higher ups in Navy forecasting.  He basically said in a perfect world my idea was a good one but their studies show the increase in model capability with air and land based info only marginally improves what the satellites provide, but sometimes it has degraded it.  I've read a little on the subject but mostly in relation to that. 

[WxUSAF]

I'm sure there's merit to it but down stream is already modeled at shorter leads with well samples UA analysis.  It's not like a vort was prog'd to hit bc but ended up at san fran. 

 

Sometimes they run recon when downstream is supportive of a high impact event. I don't think it suddenly changes anything. Maybe dials in finer details in strength and track. 

 

I might be totally out to lunch on the subject but my personal obs don't really give me confidence that all will be right with the world the second a sw hits the coast. 

It's an interesting research project to think of:  compare modeled solutions before and after the s/w arrives on the coast vs. the ultimate result, but it's of course far more complicated than that.  While the main s/w may be coming onshore, maybe the PV responsible for maintaining confluence and forcing the s/w southward may be moving northeast out of the dense North American sampling grid.  So, if a solution changes, is it because the s/w has better sampling or because the PV has worse?  Dozens of little games you can play like that. 

[Ian]

I'm sure there's merit to it but down stream is already modeled at shorter leads with well samples UA analysis.  It's not like a vort was prog'd to hit bc but ended up at san fran. 

 

Sometimes they run recon when downstream is supportive of a high impact event. I don't think it suddenly changes anything. Maybe dials in finer details in strength and track. 

 

I might be totally out to lunch on the subject but my personal obs don't really give me confidence that all will be right with the world the second a sw hits the coast. 

 

Yeah, that's true. I think the easy and clear answer is more data is always better.  But I guess my thought is it doesn't necessarily seem to really change all that much to a noticeable level at least.  2000 may be an example.. I don't really know much about that. I'll never say that CANT happen again but the increases in NWP since then are so large it's hard to see it happening.  We still get some great busts in the short term though ... I love it.

[packbacker]

Was that because it came into the west coast? Were the models that were wrong the day before better than the ones that showed us getting screwed in days prior?

 

This is what NWS said after the fact, you guys getting screwed was definitely just bad luck, although you guys had a lot of good luck earlier in 2010 if I recall correctly :-)

 

http://www4.ncsu.edu/~nwsfo/storage/cases/20101226/

 

Researchers are suggesting that the large shift in the forecast track with the 00 UTC 12/24 and 12 UTC 12/24 runs may have originated from small initial condition uncertainties/errors in the short wave trough in the central and southern Plains. Small initial condition differences with the GFS and the ECMWF at 1200 UTC 12/24 appear to have amplified to large differences in the forecast 36-48 hours later.

[WxUSAF]

After Haiyan I tried to track down the feasibility and usefulness of resuming some sort of recon (drone based).  I ended up talking to one of the higher ups in Navy forecasting.  He basically said in a perfect world my idea was a good one but their studies show the increase in model capability with air and land based info only marginally improves what the satellites provide, but sometimes it has degraded it.  I've read a little on the subject but mostly in relation to that. 

Wow...degraded it?  There can always be bad data, but there should filters to remove it in most cases.  So if it's "good" data that reduces the forecast, I'd be very interested to hear how that resulted. 

 

But back to your point about satellite data...yes, it's very good. The Global Precipitation Measurement core satellite is launching in late February.  It will be another very powerful tool that can feed into the data assimilation processes.  Not sure how long it will take to start getting GPM data into the models, but that is the intention (at least as far as NASA and JAXA are concerned). 

[Bob Chill]

Yea Wx, that's part of my logic. Subtle changes downstream have large impacts as you go out in time. It's not just sw sampling upstream. Just think of all the model comparison games we've played with 50/50's and blocking through the years. That kind of stuff makes and breaks us more than strength of a sw hitting the west coast.

I'm sure the interactions overlap in some ways but I'm not nearly smart or educated enough on the subject to say one way or another.

[WxUSAF]

I'm sure the interactions overlap in some ways but I'm not nearly smart or educated enough on the subject to say one way or another.

Nobody's smart enough.  The system is too chaotic. 

[Bob Chill]

I'm not sure the 2000 storm would bust so bad today. Models and sat data isnt perfect but it has certainly advanced quite a bit in 13 years.

If 2000 would happen today there would have been more flags and watchful eyes imo. But I'll always root for a redux with every se whiff with a pretty lp and ull.

[Ian]

This is what NWS said after the fact, you guys getting screwed was definitely just bad luck, although you guys had a lot of good luck earlier in 2010 if I recall correctly :-)

 

http://www4.ncsu.edu/~nwsfo/storage/cases/20101226/

 

Researchers are suggesting that the large shift in the forecast track with the 00 UTC 12/24 and 12 UTC 12/24 runs may have originated from small initial condition uncertainties/errors in the short wave trough in the central and southern Plains. Small initial condition differences with the GFS and the ECMWF at 1200 UTC 12/24 appear to have amplified to large differences in the forecast 36-48 hours later.

 

That storm was crazy. I don't think I can remember another trough like that in winter with how it dropped in and then was gigantic with an almost north/south orientation up the coast. In retrospect we should have known we would get screwed.

[Avdave]

0z NAM looks good thru 36 whew

[Ian]

Wow...degraded it?  There can always be bad data, but there should filters to remove it in most cases.  So if it's "good" data that reduces the forecast, I'd be very interested to hear how that resulted. 

 

But back to your point about satellite data...yes, it's very good. The Global Precipitation Measurement core satellite is launching in late February.  It will be another very powerful tool that can feed into the data assimilation processes.  Not sure how long it will take to start getting GPM data into the models, but that is the intention (at least as far as NASA and JAXA are concerned). 

Well this was in the W PAC with specific gridded settings, primarily off the NOGAPS/NAVGEM of course. Which they swear by, btw. Though I believe similar was noted in a paper on the subject from US Naval Postgraduate School which is the paper I initially started my inquest with.  Can't get a clean link without google code -- google "recon west pacific navy" if interested. 

 

I think we have enough proof through homeland stuff plus Atlantic recon to prove that the data is worth it if you can get it. I'm sure any of these issues are more with how the model ingests it or something. 

[WxUSAF]

Well this was in the W PAC with specific gridded settings, primarily off the NOGAPS/NAVGEM of course. Which they swear by, btw. Though I believe similar was noted in a paper on the subject from US Naval Postgraduate School which is the paper I initially started my inquest with.  Can't get a clean link without google code -- google "recon west pacific navy" if interested. 

 

I think we have enough proof through homeland stuff plus Atlantic recon to prove that the data is worth it if you can get it. I'm sure any of these issues are more with how the model ingests it or something. 

Thanks for the info, will try to find it tomorrow. 

 

Yeah, given that the resolution of UA sampling is still far lower than model resolutions (spatial and temporal), I still think we're in a "more data is better" situation all things being equal.