• Welcome to TalkWeather!
    We see you lurking around TalkWeather! Take the extra step and join us today to view attachments, see less ads and maybe even join the discussion.
    CLICK TO JOIN TALKWEATHER

Severe Weather 3/15 - 3/16

CheeselandSkies said:
Flip side is, sometimes it occurs but doesn't end up limiting the ceiling of the event. Based on the HRRR output for 4/26/24, I was skeptical that the northern part of the arc of warm sector storms would be able to do too much given how many were being portrayed so close together. That did indeed occur, but they ended up being prolific significant tornado producers for quite some time. I'm learning you really have to pay attention to the wind profiles to see their impact on storm mode, just having 0-3KM directional shear with high SRH values isn't enough.
Click to expand...
The thermodynamic environment on 4/26/24 was far different. It was less saturated and actually allowed for storms to vent out precipitation despite their close promximity, and i believe Fred Gossage a WHILE back mentioned that you would rather a slightly saturated profile with more dry air then a low level dry profile with even more dry air. We do not know it all about tornadoes still.

The amount of transparency we are seeing from inside forecasters to explain to the public why this went wrong is amazing. We are finally opening up to HOW events went wrong and explaining to the public where we could've done better. I seen forecaster Andrew Shearer note on X "I did not clearly communicate this conditional potential nor the failure modes so I'll try to improve on this next"

There's a bunch of meteorologists, but we're all just amateurs when we think about it. We hardly know it all about this science! Look at March 6th in Oklahoma, NOBODY expected that 5% CIG1 to verify. Yet we got a cyclic significant tornado producer. The weather just does its thing sometimes and we weren't able to message the downtrends out. We didn't really see this for 4/28/25 (speaking of, very similar analog with the positive tilt but honestly I don't think this was the main reason because positive tilts have produced major outbreaks and one in particular was 4/3/74. The main difference is it was a large, broad positive tilt with embedded shortwaves) crapvection is honestly our friend and I'm glad to see it infiltrate a major event yesterday.


I'd like to note some of the comparisons I mightve made about 4/16/11. I know this outbreak was very high end for the Carolinas, but realistically, this setup had potential to meet just a bit below that. One particular HRRR run was spitting out 2k SBCAPE, 360 0-1km SRH. That combined with any prefrontal development would be a high end run. That was before models realised the magnitude of crapvection.

My message to all of this is t take this event as a lesson in messaging and forecasting. Where did you go wrong? It's a great opportunity when these underperforming events take place. My personal thoughts is that I drifted away from my conservative approach I tried to set this year to avoid hype. I maybe fell off that train a bit.
 
KakashiHatake2000 said:
There is one I believe but I think it is used in Europe
Click to expand...
I've actually used this website before it went public because I learned about it and asked the creator if he wouldn't mind sharing the site with me. ASTORP unfortunately doesn't run on the US as it's stuck on a run from Feb 1st. It correctly predicted my squall line potential a few days ago so I was very happy with that altogether. If anyone wants to check it out, here:

 
I think another aspect of this that put the SPC in a tough position was the last minute (IE evening/night before) shift the models had in depicting the threat for yesterday. Virtually everything in the 48 hours prior was pointing to a dangerous situation regardless of whether we ever got OWS development or not. It’s rare to have the models uptrend consistently in the lead-up to an event only to back off in a major way in the final hours beforehand. This is a great example of why I think the SPC has been hesitant at-times to go higher with risk levels in the day 2/3 outlooks even when an event may look like a slam dunk.

I also want to note that we’ve had our fair share of events in the past couple of years where the models have done a poor job of modeling the timing of a 500 MB trough entering and affecting a risk area. We’ve had a few events, such as 5/20/2019, where the timing of the trough ejection was later than what was modeled and subsequently underperformed. Yesterday was another example of that. The best dynamics didn’t arrive until well after the OWS was sapped and pushed offshore.

Trey did an excellent job covering the other failure modes as well such as the lack of an EML.
 
Julian said:
I wonder if a small team of dedicated geniuses could develop AI models that issue the outlooks if it would perform as well as the entire SPC. Would be an interesting experiment.
Click to expand...

I am very confident that within a year or two we'll have AI-only models that outperform any current forecasting methods (human + model analysis)
 
lake.effect said:
I am very confident that within a year or two we'll have AI-only models that outperform any current forecasting methods (human + model analysis)
Click to expand...
I'm telling you, no matter what, AI will not pass the amount of pattern recognition. We are improving but it's not a tool that's gonna accurately predict severe weather days to the max. We will never be perfect for forecasting and neither will AI.
 
WeathermanLeprechaun said:
I'm telling you, no matter what, AI will not pass the amount of pattern recognition. We are improving but it's not a tool that's gonna accurately predict severe weather days to the max. We will never be perfect for forecasting and neither will AI.
Click to expand...

Yeah, no.

AI is already much better and faster at pure pattern recognition than virtually all humans. Just look at how large language models are starting to solve previously open math theorems.

AI is also getting faster and smarter at a near exponential rate. Today is the dumbest AI will ever be. It's only going to get smarter and more capable from here, and quickly.
 
tornado examiner said:
Well…I was right to be skeptical this time at least. I’m not upset for being antagonized over that. Cause…yeah.
But it is still notable.
Click to expand...
Now that the event is over…

You were antagonized because you’re a repeat offender. There’s a difference between being skeptical and articulating it, and what you’ve done previously on the day of events.

Your history as a whole on this forum, whether it’s been intended or not, has stirred up unneeded discourse. This spreads not only across tornado event threads, but to the EF scale discussion blow ups. Sometimes even to badgering on Twitter.

It’s one thing to be skeptical, it’s another thing to do what you’ve done in the past on here.
 
lake.effect said:
Yeah, no.

AI is already much better and faster at pure pattern recognition than virtually all humans. Just look at how large language models are starting to solve previously open math theorems.

AI is also getting faster and smarter at a near exponential rate. Today is the dumbest AI will ever be. It's only going to get smarter and more capable from here, and quickly.
Click to expand...
lol... an AI that still can't write a proper BASIC program for a commodore 64 is gonna replace humans in weather forecasting...
 
ColdFront said:
Now that the event is over…

You were antagonized because you’re a repeat offender. There’s a difference between being skeptical and articulating it, and what you’ve done previously on the day of events.

Your history as a whole on this forum, whether it’s been intended or not, has stirred up unneeded discourse. This spreads not only across tornado event threads, but to the EF scale discussion blow ups. Sometimes even to badgering on Twitter.

It’s one thing to be skeptical, it’s another thing to do what you’ve done in the past on here.
Click to expand...
Yeah…Guilty as charged officer. (Pecos Hank reference)
I deserved that

I wish I had motivation to actually study these topics properly but my mind can’t be bothered for anything anymore. I’ll try to remain as casual an observer as possible here because I just can’t add anything of importance. And not let my emotions get the better of me when it comes to opinion.
 
tornado examiner said:
Yeah…Guilty as charged officer. (Pecos Hank reference)
I deserved that

I wish I had motivation to actually study these topics properly but my mind can’t be bothered for anything anymore. I’ll try to remain as casual an observer as possible here because I just can’t add anything of importance. And not let my emotions get the better of me when it comes to opinion.
Click to expand...
Dude, you’re welcome to contribute here. I even sent you some tips last year on how to do that better. Don't be hard on yourself, we are all here to learn and we’ve all made mistakes on here. Just come here with a good attitude and put forth good faith arguments or questions. Seriously, no harm in asking questions. It’s all about how you present it.
 
“AI” can absolutely be used to aid in forecasting that human meteorologists will continue to conduct (in quotes because what we call AI today is still not truly an intelligence) but to suggest AI will take over forecasting entirely, especially in a year or two, is putting too much faith in it right now. The mathematical stuff AI has done thus far is a little overblown and isn’t super reliable yet without humans intervening and cross checking results with other sources that humans made themselves, so “AI” taking over forecasting entirely, particularly within the next 10-15 years, is just not going to happen at all.
 
lake.effect said:
Yeah, no.

AI is already much better and faster at pure pattern recognition than virtually all humans. Just look at how large language models are starting to solve previously open math theorems.

AI is also getting faster and smarter at a near exponential rate. Today is the dumbest AI will ever be. It's only going to get smarter and more capable from here, and quickly.
Click to expand...
You've been at this consistent AI regime ever since you joined and I respect you for your posts and your model outputs but I've heard this AI is getting smarter a thousand times. True. It has became tough to distinguish AI videos and that's because there's slight incorrect details still there that it hasn't got right yet.

They've said it's the dumbest it will ever be for two years. I don't see exponential growth but small steps. Forecasting is not a perfect thing and it never will be. I'm sorry but this is just plain out AI hyperbole they've been promising.

AI is literally used half the time for fake "Chihuahua rides cars" videos on Facebook because the platform's degraded to slop. And that's what you expect to overtake. Please stop. There's so much variables in weather forecasting AI will not be capable of doing. And if they do, I'll put my words back up my wahoo but I just don't see this vision of AI replacing human forecasting for a while. It can assist but i wouldn't let it do the work for us. That's my general opinion.
 
tornado examiner said:
Yeah…Guilty as charged officer. (Pecos Hank reference)
I deserved that

I wish I had motivation to actually study these topics properly but my mind can’t be bothered for anything anymore. I’ll try to remain as casual an observer as possible here because I just can’t add anything of importance. And not let my emotions get the better of me when it comes to opinion.
Click to expand...
You've kinda said this same "I'll try" thing for the past few years and then you continue to result to the same "trash or bust" thing when a event comes around. You're welcome to contribute and it's appreciated. If you were motivated, I'm sure you'd bring greater results to the threads. You just have to take it easy sometimes.
 
You must log in or register to reply here.
Back
Top