DTP

Agreed- however, we live in America- I believe the concept of possible v probable for most people has become buried in delusion and pot gummies lol

Think my new rule will be 72 hrs or less after this- I did more research into the math/assumptions/calculations behind the Euro and GFS yesterday; overall I feel that most of the variables are satisfied appropriately, and the overall construction of both is very robust- I would feel confident in receiving accurate information from them (all things being equal) when they receive the correct data. However, that is also the biggest flaw outside of 72- really not their fault tbh (the calculations are correct and the assumptions are historically realistic)....problem is the with the data, and will continue to be. The only error they make is in making long term prediction off of spotty and/or hypothetical downstream data. Classic inference/extrapolation fallacy, which makes a robust model really thin past a certain point of time. If you think of it like normal distribution- you are using data from the tails to forecast the mean when you go out past 72 hours or so It is backassward, but not intentionally....you gotta have the right tool for the job Beavis

Lol- yessir Unfortunately, now it becomes a "boy who cried wolf" scenario- confusion amongst the public = them not taking this seriously because of a mixed message Case in point- yanked my gennie out in the driveway yesterday evening to check fluids and make sure it was running properly- had 3 separate neighbors come over and ask why I was running it, so I mentioned that we may have some outages...two of whom said "I heard on my weather app it wasnt supposed to snow"::face palm::

100%- count me in lol In light of this total clusterf$%k, they need to move their prediction scope backwards to 72 hrs max on just about all of these models ouside of a very basic probabilistic outcome (with a wide safety net) until they ingest metadata with a closer linkage to an actual event. The flight into the Baja low last night confirms this Some of the models I have built are for US-based equity futures and futures options (SPX/NDX/RUT)- if I were to have to made this magnitude of an error, someone(plural) big goes bankrupt, especially since it involves delta hedging by prime dealers (market makers). I know I sound like a broken record about incorporating metadata, but my models are comprised of approximately 60-70% (or more) of a combination of real-time/near-term. Market makers have to stay delta neutral (first order Greek) at all times, and also account for the four other second-order Greeks in Black-Scholes; plus inventory (sold calls/puts, often in complex spreads) has to be zeroed out constantly. And a lot of the flow is HFT (high-frequency algos)....and my models have to predict this 2 to 3 days in advance. Why not more? Because the variance/covariance and correlations completely lose any accuracy without recent metadata- they go from a very robust model to stick-thin with no predictive power within the span of two trading days ....the phrase "you can't bring a knife to a gun fight" comes to mind

Well, guess we are getting somewhere now....j#$sus what a mess if it verifies- going to have a lot of people stuck in their neighborhoods where I live with no power....ZR on top of IP is impossible to drive on, up or down a hill without snow tires or chains. Having 4wd doesn't make any difference in that situation

Ahh- I see....and yes, that would change pretty much everything at this point...early phasing would act as the kicker on the 50/50 and make it more progressive, correct? Some of the ensembles were showing this, but then I saw some others that had some genuinely bizarre solutions I still am skeptical of this without seeing more of the metadata being fed in- many of the other ensembles were all over the place with no coherent line of best fit

Could very well be the case- or tried to correct upon multiple variables at once (shotgunning). Or corrected for another completely separate dynamic that typically does not influence output, but did in this specific case, or the solar storm (plausible, hardware, communications and electronics in general do not respond well to magnetic waves) Or could be sniffing out something extremely early I would take any of what you see on the models with a grain of salt until metadata is derived from the low and actual surface reports- much cleaner picture

Haha- yes, not a thing, never has been, never will be....all models are showing 020-050 degree wind headings throughout the storm at CLT. They do not deviate much, if any, between models. NE wind = CAD from a 1040+ high over NY state throwing out dew points in NC/SC down in the teens...using this logic alone, if the low plowed through the wedge, the wind field would veer out to ENE or even south of 090.....then the CAD would become in-situ or not even present in the first place Also why they are looking at the actual metadata from the Baja low before buying into it. Metadata has already occurred, so it isn't influenced by any of the models. When the models begin to feed in the metadata, you will more than likely see several of them singing a different tune.... They obviously have some good forecasters over there who have a good handle on the potential flaws in the Euro- it takes a lot of experience to come to that conclusion after picking the ensembles apart

Good catch- and there is a lot of truth to the KFFC's assessment; it feeds into what I was saying earlier about probabilistic models after having worked with and creating them....you have to discount data at times when it simply does not make logical sense given the situation

I will do a video behind my house- we have a golf course running to the bottom at the lakes and a driving range on the other side of the hill. The driving range is on top of the 150' hill with a 30 degree pitch- it more resembles a ski slope than anything else The neighborhood kids use river tubes and sleds- somewhat uncertain how someone hasn't gotten killed at this point- they build jumps in as well....some of them hit 40 mph+ otw down

2.4" of total precip @ 4:1 would be ~6" if the precip were all IP....which would take about 9 centuries to melt off with as cold as it is going to be Most I ever saw was 3"+ in Dallas; it was back to 70 degrees 4 days after the event- really strange seeing the ground still coated- it stuck around for almost a week after 10" at 4:1 is insane- that system was juiced for that to happen- 4-4.5" total QPF

Having lived around CLT (Highland Creek, Harrisburg, and Tega Cay) for 25 years, I have seen the same setup quite a few times at this point. If you are in CLT and want 100% snow, it probably won't happen- very low probability, even north of 85. If you like sleet. and lots of it- this is your storm; it will be the predominant p-type south of 40 down to a Chester/Union/Greenwood line. South of that, mostly ZR down to the coastal plain. The depth of cold air is extremely impressive for a CAD event; this has not deviated substantially on any of the models. Yes, a warm nose can and will wreak havoc on a pure snow event, but it is always a big feature in dammimg events- no way around it in most of these scenarios. However, if you have aome sort of frozen atw down to Wilmington and MB, and have a strong CAD signature going to NE Alabama, you are going to have high-impact winter storm from I-20 north. N Texas has different topo than the Carolinas, but the manner in which cold moves southward in an event like this is somewhat similar to a CAD event in our region (I grew up 20 min north of Dallas). Deeper cold layer the further north you go, shallower as you go south from the Red River. Backdoor front with NE winds during CAD is fairly similar to a blue norther with attendant instability and p-types as the cold filters in, especially with how the track of this low is modeled. So you have WSW almost down to Austin at this point- that is a much deeper layer of cold around Dallas (upper teens by late Saturday). Start as rain, then fr rain, then goes to a mix of frozen with predominantly sleet, and quite a bit of sleet from what I am seeing. Most likely the same scenario in and around CLT, as our modeled temps are pretty similar in this setup. Thinking maybe 2-3" of snow in CLT, 3-4" of sleet, and minimal freezing rain- sure, you can have freezing rain in the upper teens/low 20's here, but it isn't likely given the amount of cold air present. Biggest ZR amounts will be just north of 20 in SC over to outside of Wilmington. 40 north probably sees 6"+ snow totals, plus a good bit of sleet....prediction on my end is based on living in the region and having seen this setup several times a decade....it isn't all that common- especially with the magnitude of cold present Fwiw- metadata beats any probabilistic model 100% of the time (I have worked with probabilistic models for 20 years in health science and finance, as well as created them), simply because metadata has already occurred- so take a look at METARS totals from the DFW area starting Friday PM, and you will have a really great crystal ball as to what is going to play out in the Carolinas, at least around Charlotte. I wouldn't hold much stock in any model run this far out, outside of the overall probabilitstic signal (major CAD/winter storm for the region)- there will be some major fluctuations- lot of noise, wo specifics....L/T AI will generally improve predictive value on the majors, but it has to learn first. Also a testament to how fantastically bad some of these models actually are to start out with- when you have almost unlimited computational power, much of what you are seeing in terms of run to run discrepancy should have already been smoothed out by now to some extent. I have seen rapid smoothing and a huge increase in model robustness in industries with an even larger variable set and more complex scope of data, simply bc the existing models gave AI a much better starting point..... Good luck to everyone on this, and make sure your loved ones have access to heat and electricity/generator. Also have a gutter company on speed dial- a 5" gutter spike into fascia will not hold up with that amount of sleet/ice over a 25' run...I have been meaning to put up gutter guards for about a year- I think I will get the ladder out before Friday and get it done