Jns2183

The 19z NBM probabilistic snowfall guidance for the Saturday night through Monday morning window shows a really sharp divide across the region. It's not just "big snow might happen"—the risk profiles vary a ton depending on where you are. For spots like Lancaster (KW05), Pottsville (KZER), and York/Gettysburg (KTHV), the floor is nasty: their 10th percentile sits around 10–12 inches. That means even if things underperform noticeably, you're still looking at enough snow to break out the shovel and measure in proper accumulations. That's typical when the cold air is locked in and the big question marks are how long the heavy precipitation band sticks around and exactly where the deformation zone or mesoscale banding sets up. Over toward Harrisburg/Carlisle (KCXY) and the Reading/Philly corridor (KRDG), it's the classic high-end skew: medians in the 14–16" range, but the 90th percentile climbs to 32–35" or so—more than double the median in some cases. That's straight out of the playbook for those arctic, high snow-to-liquid ratio setups. The outcome rides almost entirely on mesoscale features: where the strongest frontogenesis band parks, how long it persists, and whether lift really peaks in the dendritic growth zone. A tiny shift in that band can turn a solid warning-level event into something historic at one spot while the next town over stays more moderate—all from the same synoptic storm. Up north at places like KZER and KWBW, the spreads look tighter and more stable. That tracks with a solidly entrenched cold dome: minimal risk of mixing or sleet eating into accumulations, fewer p-type headaches, and totals mostly hinging on how long the precip keeps firing rather than ratio roulette. So the southern and valley areas are wrestling with the full uncertainty beast (track wobbles, banding details, ratios), while the ridges and northern zones basically boil down to one main question: how long? The 07z to 13z to 19z NBM runs on January 21 show the storm strengthening and coming into clearer focus rather than falling apart. QPF keeps climbing steadily across the board with each cycle, while snowfall amounts spike sharply from 07z to 13z and then basically level off near their highest values by 19z. That's a classic sign of the models gaining confidence: the debate shifts from "does this even happen" to "where does the heaviest axis land and how efficient will the snow be." The 13z run had the fluffiest ratios—some spots like State College and Wilkes-Barre pushing 20:1 or better—but 19z pulls them back to a more realistic 15–19:1 range. Totals hold strong, though, because the extra moisture more than makes up for the slightly lower ratios. No major backtracking shows up late in the day either; places like Lancaster and Harrisburg stay nearly flat from 13z to 19z, which is reassuring as we get closer. The northern edge (KWBW) sees the biggest jump, with totals climbing from 7.8" to around 14" and holding there, suggesting the heavy band is expanding farther north. Overall, the cold air looks solid, moisture is trending higher, and the guidance is locking in on a legit high-impact snow instead of still questioning if it'll show up. Sent from my SM-S731U using Tapatalk

These images may be of help Sent from my SM-S731U using Tapatalk

Here's substack post from new app https://open.substack.com/pub/actuallyweather/p/which-forecast-is-best?utm_source=share&utm_medium=android&r=16mod Sent from my SM-S731U using Tapatalk

For anyone who wants a look at Google's new weather model, meteorologist Eric Fisher has a simple forcast viewer on his github https://efisher828.github.io/weathernext/ Here's more information on it https://deepmind.google/science/weathernext/ https://developers.google.com/earth-engine/datasets/catalog/projects_gcp-public-data-weathernext_assets_weathernext_2_0_0 If anyone has the technical background I believe you can already do a whole lot more Finally this new endeavor I came across that has a whole lot of potential even though it currently only had temperature and wind verification scores. https://www.actuallyweather.com/?run_date=2026-01-21 What's neat is it has it by major weather reporting station, like kmdt. The promise is soon we could have scorecards for about 50 variables like snow, and rainfall, along with ability to customize periods. And yes, I understand the scorecards put out by the nws and euro are for much larger areas and maybe even tracking slightly different metrics, but dear God they require hours of study to even interpret probably as the user base for them usually have phD. While I understand that accuracy in temperature or qpf for a single location has the probability of noise being potentially exponentially bigger than the signal, (i.e a model, especially an mL one that continuelly maximizes minute station details may quickly Decoherence at the upper air forcasts that drive accurate forcasting at a greater geospatial scale; in happy that a true effort is finally being made to score at the level we all care bout and spend so much effort discussing. Sent from my SM-S731U using Tapatalk

Here's from an infograph I made from the text bulletins of the 19z NBM. specifically regarding snowfall percentile amounts by region Sent from my SM-S731U using Tapatalk

Here's some maps of this area's most famous cold sleet storm, because if it's going to sleet it's going to do so into some cold surface air Sent from my SM-S731U using Tapatalk

I wish for a twitter battle between him and the WPC Sent from my SM-S731U using Tapatalk

If anything this reminds me of December 2009 Sent from my SM-S731U using Tapatalk

I don't think you're going to know till within 24 hours with how models have been. Sent from my SM-S731U using Tapatalk

@MAG5035 please correct me if I'm mistaken. I could totally be off on a wrong track and have no idea. I'd rather be wrong and learn than keep spouting crap. Sent from my SM-S731U using Tapatalk

I think this storm is partly here due to la Nina breaking down. That increases volitilty significantly which is what we want for a big storm. The down side is the northern trend all winter has been due in part to the la Nina. I also feel, and I believe, studies have panned at that the northern trend that models do well above 50/50, isn't do so much to the teleconnection as it is data availability and it being much more available at lower latitude. If anything the prudent choice of action would be to see which models have had the better accuracy with high pressure placements at higher latitude both all winter and during the past month as that seems to be the hinge this storms depends upon Sent from my SM-S731U using Tapatalk

Over/under on. 5 tweets this week? Sent from my SM-S731U using Tapatalk

What's the Lancaster guy saying? Sent from my SM-S731U using Tapatalk

Are those by days of snow individually or do you compress multiple days to one storm. If you want I can send you how I did Harrisburg as well as raw data going back to 1890 Sent from my SM-S731U using Tapatalk

At day 5 compared to day about 60-70% of the low mean pressure has come north with a mean distance of 100 miles Sent from my SM-S731U using Tapatalk

This winter like most winters I would prefer to be on the northern fringe until 36-48 hours beforehand. This winter it's gone down to 12-24 hours. Most of our huge storms have followed the same pattern. Tonight and tomorrow I'll see what I can dig into regarding forcast days in advance and north trends just to confirm it isn't confirmation basis. The other interesting feature I glanced at was that the swing from bad luck to good luck seemed more often than not to be a pretty big storm. I want to put some quantitative numbers to that though. Sent from my SM-S731U using Tapatalk

Best position since 2016 Sent from my SM-S731U using Tapatalk

I remember laughing out loud when after all the damn hours I spent collecting daily data for NAO, PNA, AO, GBI, MJO, a couple other indicies going back to 1950 and running all the multivariate time lag correlations that at most those indicies explained roughly 6% of snowfall variance in our area. Further, that if I could have a super computer that held infinite historical data it would be a miracle if I approached 40%. So much of our storms come down frankly to dumb luck and pulling the slot machine handle. That said it isn't a slot machine, and our luck the past 5 years has rivaled the 1950s for worst luck ever. Eventually the luck will change. This weekend is as good as time for it to change as any I've seen. Sent from my SM-S731U using Tapatalk

Apparently the neighbors who got popped had a multi year fued. I remember reading up on it and it was so ridiculous I was in shock nothing happened sooner. I believe it's worth stressing the backstory here, else people believe they are at risk from getting it from any of there neighbors at any time. Not that they deserved this, just that your chance of being gunned down by neighbor are in someway usually proportional to your behavior over years. Sent from my SM-S731U using Tapatalk

I made a big one and he dove in head first before turning around with just his head sticking out and stayed in that position content for the next ten minutes Sent from my SM-S731U using Tapatalk

Spent an hour outside with him happy as can be Sent from my SM-S731U using Tapatalk

Any chance you have the original uncompressed images for the first 5 Sent from my SM-S731U using Tapatalk

I have close to 3" of snow. I also picked up this little guy yesterday. This morning he was scared of snow until he got in it. Then immediately went crazy with delight Sent from my SM-S731U using Tapatalk

Thought people might enjoy this The temperature is based on recorded max temperature for day and the percentage is out of all the days preceiptitation was measured Sent from my SM-S731U using Tapatalk

Final part Haha, I spent way too long on this to not put forth my pet theory regarding the 1960s. The absolute insane return periods for "luck" and just plain actual snow and it's correlation to some events has not been studied super well. I guess this is the sacrifice needed for us to get the winters we pray for. Sent from my SM-S731U using Tapatalk