#philly #weather #pawx #philadelphia #phillywx #snowstorm #blizzard 2026
An archive of the forecast posts leading up to the Blizzard of 2026 in Philadelphia
Was the snow storm really a ‘Blizzard” as defined by NOAA? A review of the NWS KPHL website shows only about 1 hour visibility of 0.25 mile or less, instead of the defined 3 hours. Wind gusts were not sustained at over 35 mph for 3 hours either. So, while the storm will likely be remembered as the “Blizzard of 2026”, it really wasn’t. (Areas near the shore likely qualified for the blizzard designation.)
A few things about the models. Timing the transition from rain to snow was very good. Perfect snow total predictions are never achieved, just as rainfall is never exactly as forecast.
It should be remembered that the GFS model did much better than the ECMWF with this storm, for days in advance. The ECMWF-AI did much better than the regular ECMWF. The Canadian models were under-performers as was the German ICON model.
The new RRFS, waiting to be released, was disappointing, especially with its forecast at 60 hours and 84 hours in advance. However, the 24 hour snow forecast was very good.
The old NAM did much better than the RRFS at 60 to 84 hours. The NBM median was pretty good, maybe a tad high in some areas.
Snow Tapering Off
Posted Monday 02/23/26 @ 9:51 AM — The storm is well to our east and the vertical motion (Omega) near and west of Philadelphia has turned to a downward motion. Snow continues to rotate around the low into the city from the northeast along the 700 mb (~10,000 ft) wind direction but as it meets downward vertical motion, it should taper off. Additional scatted snow showers still possible into early afternoon near the city.
Additional snowfall from the 13z NBM (8AM) run is minimal west of the city.
I’ll do a wrap up and forecast review later today.
Additional Snowfall
Posted Sunday 02/22/26 @ 10:27 PM — Despite the afternoon’s models suggesting a more eastward track, the latest model blend (NBM) shows significant additional snow from 8 PM through 4 PM. (In the city, snow tapers to just snow showers/flurries by 12 PM)
If this holds, the NBM mean snowfall (appeared too high) did the best at predicting this storm’s snowfall.
Updates Monday morning.
Storm Update
Posted Sunday 02/22/26 @ 5:58 PM — Folks, I violated my own cardinal rule- “never look at the new model runs when the storm is in progress”.
The reason: it always changes and you never know if it’s a dramatic change in the forecast or if it’s a model “spin-up” error. Put another way, the latest models are not often not the best once a storm is in progress. I mentioned this aspect this this morning.
So, I looked at the latest GFS and RRFS models (18z) , and I shouldn’t have been surprised that the predicted snow amounts are significantly less. Is this a real change? Or a spin-up error?
(Models literally re-create the atmosphere mathematically; they often have a multi-hour “spin-up” time for the model to come close to matching the atmosphere.)
Anyhow, here’s the latest RRFS, which has cut back the snow totals. (The GFS has gone even lower to only 6 inches in Philly.)
Snowstorm Update
Posted Sunday 02/22/26 @ 3:51 PM — As forecast, it started snowing between 3 and 4 PM in many areas. Near ground temperatures are still above freezing, according to the RTMA, (however the hourly RAP model indicates that temperatures about 800 feet above ground are at or below 32º) —
The 1 PM (18z) HRRR has chimed in with its snow forecast and it remains in the range I’ve been supporting—
Last but not least, the latest NBM run with updated precipitation (19z) is available and the median snowfall has increased from the last run. I’m still placing bets on the median snowfall. (hard to believe, but the mean is even greater)
Sunday Storm Update
Posted Sunday 02/22/26 @ 11:50 AM — The latest RRFS finally caught up with the rest of the models regarding snow accumulation—
Additionally, a newer run of the NBM (13z) showed somewhat higher median snow totals than what was posted earlier.
No real change in the timeline. Snow starts accumulating between 2 and 4 PM RRFS model—
Winds
Sunday Storm Update
Posted Sunday 02/22/26 @ 8:28 AM — The models show little change from last night, perhaps a bit higher snow totals. The ECMWF’s low pressure position finally caught up with the position the GFS was forecasting all week. (Who says the ECMWF is the Mercedes and our GFS is a Chevy ?) Here’s the current position of the low—
Time line is similar to that posted yesterday. Current rain or rain/snow mix changes more to snow around noon or 1 PM. Then all snow by 3-4 PM in the city and westward. Accumulations start about 4 PM or so. The storm tapers and ends around 11 AM- noon Monday. Some additional snow showers possible in the afternoon.
Here are some of the latest snow accumulation totals—
NAM built-in algorithm— (Generally too low)
The NBM median and mean (implying higher certainty) are reasonably close together and I’m going with the median snow accumulation below (the mean values are even higher!) —
The GFS built-in snow depth values are about 1 inch less than the the NAM shown above.
Here’s my “experimental” values calculated on GFS data of precipitation and temperatures at 4 levels and water-snow ratio; they are similar to the NBM median—
At this point, with consistent model forecasts, I’m not inclined to change my forecast. Any newer models today may not have the proper ‘spin-up’ time to capture the atmosphere.
I know my snow forecast with this storm is different and higher than forecast by the TV forecasts and the NWS.
Interestingly, the RRFS and REFS (expected to become operational soon) are in the 8-10″ range for the city. I’m somewhat disappointed that both models are slow to pickup this storm at 60 and 84 hours and are perhaps forecasting snow totals. that are too low, especially the RRFS.
I guess we’ll see what happens.
I’ll update during the day as needed.
#Philadelphia #weather #PAwx #PhillyWx
Sharing Excellent Question/Comment Submitted
From Brian Friday, February 20, 2026 10:40 am at 10:40 AM
Thanks for sharing your love for weather with the broader community. It’s always fun to hear your perspective on the raw data.
Just curious: do you ever peek at the NWS forecast? Other than you 🙂 I think of them as the gold standard. They’ve been consistently forecasting on the lower end for my area–shocking close to the blend median.
Also I’m curious if you’ve focused on the air temp during the bulk of the precipitation. In my area, it is forecast (who knows what really will happen!) to be decently above freezing–in the upper 30s. Is that a factor in the argument for lower totals? Does that suggest that even if we get a decent snow, a lot of it will be on grass and not roads and sidewalks?
Just trying to get a handle on the practical implications for whatever is coming.
Hi Brian, Excellent questions and comments. Let me address them—
1: NWS: I am totally impressed with the forecasters and scientists of the NWS. And I DO more than peek at their forecast discussions on a regular basis. But I find myself unduly influenced if I read their forecast discussions before I make my own conclusions.
Their forecasts have legal weight in case of public injury and destruction and I find that they need to protect people and property first, often erring on the side of over-stating certain forecast risks. Severe Thunderstorm warnings are the perfect example. When it comes to snow storm risks, they also focus on the economic impact of a wrong forecast. So they don’t move towards excessive snowfall extremes until they have to.
As you’ve seen, I’m currently inclined towards the NBM median but it may be too low.
2. Temperatures and Snow: I’m totally focused on temperatures with snowstorms. But near-ground temperatures (actually 2m above ground is what the ground temperatures really mean) are meaningless when the upper level temperatures from 1500 feet to 10,000 feet are below freezing. (It will snow 50% of the time when near ground temperatures are as high as 36.5º and the temperatures from 1500 feet to 10,000 feet are at or below freezing.)
The red, magenta and yellow lines on my maps are “critical thickness” lines which show the freezing temperature at various levels (actually thicknesses) of the atmosphere. If all three lines are south of your area, temperatures aloft are cold enough for snow even if it’s above 32º near the ground. Where ground temperatures come into play is whether the snow accumulates, compresses and/or melts. Model algorithms try to take that into account.
Interestingly, I was going to unveil a new algorithm I worked out over recent weeks with Claude AI for GFS snowfall totals. I will run this algorithm on the GFS data I’ve scripted to download and process 4x daily. I may try it with today’s 18z GFS totals. Temperatures aloft are a major part of this algorithm, as is the temperature near the ground. (I’ve set the cut off to be 33º, BTW.)
Thanks for your questions. I may share it and your response in the main blog.
Storm Update
Posted Friday 02/20/26 @ 10:02 AM — There has been some improvement in the forecast, as far as the NBM is concerned. But there are general trends and differences that remain—
- The ECMWF maintains a weaker storm until it passes us by. It’s track has moved somewhat closer to us, but still low snow totals are forecast (1-3″)
- The GFS continues to forecast a major storm with snow totals over 6″ in our area and more at the shore.
- Disturbingly, the NAM has the energy jumping over our area, keeping the snow totals very low.
- Similar to the NAM, the RRFS shows very low snow totals with a far southern track.
- The Canadian RGEM and GDPS AI area all in the low accumulation camp.
Here’s the ECMWF forecast at 1 AM—
The GFS maintains a monster of a storm with heavy snow in our area—
The other models have similar to slightly less snow accumulations as shown in the table from yesterday.
Here’s the latest ECMWF-AI forecast—
At this time, I’m going to depend on the Model Blend (NBM) and I’m leaning towards the median snowfall forecast. Here’s the latest—
Here’s the latest NBM mean forecast—
The median and mean have become much closer lending higher confidence. But I still can’t get past the NAM and new RRFS that show minimal snowfall here.
Updates this afternoon.
Snow Storm Update
Posted Thursday 02/19/26 @ 8:54 PM — The latest ECMWF has moved in the direction of the GFS, as has the latest ECMWF-AI and the AIGFS. A major snowstorm is being predicted by all four models.
The latest ECMW-AI snow accumulation forecast—
I’m not ready to share the GFS totals. It’s either wrong or we’re headed for a whopper of a snowstorm.
We’ll have to see if tonight’s model runs are continuing with the current trend. Stay tuned!
Sunday Storm Update
Posted Thursday 02/19/26 @ 4:48 PM — The forecast for the coastal storm on Sunday continues to evolve. Here are some of the latest trends—
| Model | Current Forecast Snowfall | Comments |
| ECMWF | 2-4″ | southern track |
| ECMWF-AI | 1.5-2.5″ | southern track |
| GFS | 6-11″ | Northern track Phased |
| AIGFS | 6-12″ | Heavy at NJ shore |
| HGEFS | 2-5″ | NJ shore heavier |
| GEFS | ||
| GDPS AI (Canadian) | 3-4.5″ | |
| ICON (German) | minimal | southern track |
| RRFS | minimal -2″ | southern track |
| RGEM (Canadian) | 0-2.5″ | snow west of PHL |
| NAM | 0- trace | southern track |
| NAM-NEST | N/A | not in forecast range |
| REFS | N/A | not in forecast range |
| HRRR | N/A | not in forecast range |
| HRDPS (Canadian) | N/A | not in forecast range |
| NBM mean | 7.7″ | latest 19z |
| NBM median | 2-3″ | latest 19z |
| NBM 75 percentile | 9-16″ | latest 19z |
There you have it. Clearly, the NBM model blend is highly influenced by the GFS.
I’m leaning towards the NBM median at this time:
Of interest is the wide disparity between the GFS and the NAM. (I dusted off my download code to retrieve and process the NAM. The NAM will be retired this year.)
For many years, I relied exclusively on the GFS and NAM for my snow forecasts. What I distinctly remember- the NAM tends to over-forecast snowfall and an average of the NAM and GFS usually served well when their forecasts were close. But, when the NAM is forecasting a miss (as is the case here), the lesser NAM forecast was the more accurate model.
The GFS has undergone many revisions since this observation. The NAM hasn’t been updated since 2017. So that past experience may not hold.
Latest Thoughts on Potential Sunday Snow
Posted Thursday 02/19/26 @ 9:30 AM — As last night’s Quick Update indicated, several models which leaned towards the southern track (missing us) moved back towards a somewhat more northern track. This includes the ECMWF-AI, GFS, German ICON, Canadian GDPS-AI Spectral, HGEFS, AIGFS. and GEFS.
The latest ECMWF maintains a southern track, but shows some light snow in our area with the upper air low. The ECMWF has been the most consistent which earns it some respect in this forecast puzzle and is the least “phased” system.
The latest GFS has moved back to a forecast of substantial snow possible—
With the ECMWF’s consistent southern track, it is worthwhile to see what the statistical ensemble of the ECMWF shows regarding surface low position and location of most uncertainty—
The mean position of the low in the ensemble is more northern than the regular ECMWF and the greatest spread of the ensemble members is towards the northwest of the mean position, suggesting a strong possibility of a more northwestern track.
So we can’t ignore the ECMWF, but there’s too much uncertainty within the model variants to assume it to be correct.
Later today, the storm time horizon is coming within the forecast range of the NAM, RRFS and Canadian RGEM (all forecast out to 84 hours.) We’ll probably have a better handle on the forecast at that time. Updates later today.
Note— Most updates on this site are cross-posted on social media. To save time, posts labeled “Quick Updates” are not.
Quick Update
Updated Wednesday 02/18/26 @ 11:34 PM —Tonight’s GFS has returned to forecast of significant snow lasting into Monday. Updates tomorrow.
Uncertain Sunday Storm
Posted Wednesday 02/18/26 @ 5:26 PM — Over the past day, we’ve gone from a potential whopper of a snow storm to a highly uncertain forecast. With the advent of AI modeling, it had been my hope that this sort of great uncertainty with coastal storms would be a thing of the past. Apparently, the AI models are just as fickle as the older numerical models.
Regarding the storm forecast for later Sunday, the ECMWF maintains a southern track and suppressed development. The latest available 12z run shows it totally missing our area—
The latest GFS has reversed some this morning’s trend towards the ECMWF forecast and now shows a storm capable of greater than 6 inches snowfall—
The latest ICON model was closer to the ECMWF and the latest Canadian Global Spectral AI was in-between the GFS and the ECMWF.
The latest ECMWF-AI model is closer to the GFS than it’s own regular ECMWF.
So we have a highly uncertain forecast at this stage with snowfall forecasts ranging from zero to over 6 inches.
The storm is still over 110 hours in the future. We’ll need to wait for the event to come into the range of 84 hours for some input from the other higher resolution models. Stay tuned.
Sunday Storm Update
Posted Wednesday 02/18/26 @ 12:33 PM — It’s interesting that the lone outlier model (ECMWF) in yesterday’s cache of medium range models was actually leading the pack in forecasting a southern track and a somewhat delayed intensification of Sunday’s possible storm. The list of models forecasting a major storm have moved in the direction of the ECMWF today. Here’s the latest GFS—
There’s still plenty of time for things to change, but a southern track would be most consistent with the storm track pattern we’ve seen in recent weeks. Stay tuned.
Updated Wednesday 02/18/26 @ 6:48 AM —A quick update on Sunday’s potential storm. Last night’s models have moved towards the ECMWF with a more southerly and easterly track with less development offshore. Still stormy but much lower potential snow accumulation (4 inches or less). I’ll update later this morning with the new models runs (12z runs).
Weather Update: Wednesday-Sunday
Posted Tuesday 02/17/26 @ 7:21 PM — A warm front will move through Wednesday. Temperatures will reach the low 50s, but with the warm front, light showers and drizzle in the morning, some additional showers are expected in the mid to late afternoon. There may be some breaks of sun around noontime.
Following a weak cold front with high pressure to our north, an easterly flow on Thursday will drop the temperatures down with plenty of low clouds.
Another rather wet system approaches on Friday with rain for much of the day. A continued easterly flow will keep temperatures in the 40s.
Saturday looks cloudy in the morning, sunshine in the afternoon. Temperatures in the upper 40s.
Most models are predicting a deep coastal storm to form on Sunday. The exception is the ECMWF, which keeps the storm to our south.
Unlike the the past system that brought us light snow this pastSunday night, (and whose prior forecasts kept changing), this Sunday system is is being consistently forecast.
It seems that less a question of whether the storm will form, but whether it will be snow/rain or a mix with near-ground temperature forecasts uncertain. That said, many models are forecasting 6-12+ inches of snow at this time with the storm lingering into early Monday. I’m not ready to hang my hat on these high snow totals.
Here’s the latest GFS forecast for Sunday—
The following models currently have a similar forecasts to the GFS above— AIGFS, ECMWF-AI, Canadian Global AI-spectral, NAEFS, German ICON, HGEFS. They differ in temperature forecasts and have slight differences in track and intensity. An interesting system that bears watching, especially with the ECMWF being the exception.
Expect additional clouds Thursday with an approaching rainy system on Friday.
While it’s a long way off, the ECMWF-AI shows a potential snow event for Sunday.
I expect the forecast to change as the week progresses. Stay tuned.
Final Storm Thoughts for this Evening
Updated Saturday 02/21/26 @ 11:40 PM —Based on tonight’s models, I’m thinking the TV forecasts of 8-12 inches of snow near the city may be too low. I’m staying with the NBM median and even my experimental forecast below.
Posted Saturday 02/21/26 @ 8:51 PM — The snow totals forecast I heard on TV this evening was 8-12 inches. A reasonable number, and depending on how much precipitation falls before 4 PM may make that the correct forecast. However, the latest NBM still shows higher median snow totals as shown earlier.
My GFS snow experimental forecast is much higher and may very well be wrong, especially if much precipitation falls before 4 PM. But here’s the latest in what I’m calling ‘experimental’ forecast based on GFS data —
My forecast above isn’t very different than the latest NBM median—
The real time weather analysis (RTMA) shows low pressure just beginning to form off the coast, as picked up by 3 hour barometric change—
Final thought for tonight. Snow during this period in February melts quickly. Sun angle is high and even with clouds, radiant insolation is strong. It’s not expected to be frigid afterwards like after the last storm.
Sunday Snowstorm Update
Posted Saturday 02/21/26 @ 4:46 PM — Let me address a few things that were not given attention to.
Current water vapor shows the storm beginning to form near SC coast—
Precipitation starts as either light rain or light snow or a mix as early as 5 AM Sunday near the city.
- Despite falling as snow by 1-3 PM, little accumulation occurs until 4 PM
- Winds pick up about the same time. Winds 20 mph gusting to 40 mph.
- Precipitation ends around 12-1 PM Monday, later somewhat in eastern NJ.
The latest NBM (19z) median snow totals are below. The mean totals are much higher. I’m still leaning towards these values.
Where could these totals go wrong? The RRFS has a good portion of the precipitation falling during Sunday but not accumulating. Is it a model error or does it know something the other models don’t? Here’s the RRFS time line—
I’ll update this evening if things change.
Yet Another Storm Update
Posted Saturday 02/21/26 @ 11:41 AM — The latest GFS just became available. Here’s its built-in snow depth parameter (which I find to underestimate snow accumulations)—
Here’s what I’m calling my “experimental” snow accumulation estimate, using my own algorithm calculations on GFS data—
Updates later this afternoon. Check back.
Winter Storm Update
Posted Saturday 02/21/26 @ 10:17 AM — Additional models have become available. A picture is worth a thousand words—
Timing update: Light rain and or snow begins as early as 4 AM Sunday. By 9 AM, the latest RRFS shows this precipitation type:
Precipitation falling as snow does not equate to accumulating snow. Temperatures above freezing with high sun angle and insolation through clouds in late February will likely reduce or eliminate snow accumulation before late afternoon or evening in many areas.
Updates later. Stay tuned.
Winter Storm Update
Posted Saturday 02/21/26 @ 9:24 AM — The models still show some differences in snow totals. What had looked like a uniform move to the higher end last night has splintered again into the lower totaling ECMWF/ECMWF-AI and the NOAA models GFS/NAM/HRRR which are forecasting higher amounts. (For reasons to be determined, the new still unreleased RRFS is in the low end of about 5-7″.) The latest NAM, NBM and others will be available over the next hour or two.
Much, but not all of the storm, now falls into the range of the high resolution HRRR.
The latest HRRR (whose forecasts only extend to 48 hours) includes a part of the storm up to 7AM Monday—
I’ll be making additional updates before noon. Stay tuned.
Storm Update
Posted Friday 02/20/26 @ 10:15 PM — A quick update. Tonight’s models have moved towards higher snow totals. The NBM median is almost as high as the mean.
The latest NAM is even higher. Updates Saturday morning.
https://www.meteor.iastate.edu/~ckarsten/bufkit/image_loader.phtml?site=kphl
Experimental
Posted Friday 02/20/26 @ 5:51 PM — With the soon to be retired NAM, I’ve been working on a snow algorithm with considerable help from Anthropic’s Claude AI that uses my snow forecasting technique derived from the old NAM FOUS data that served me so well years ago. I’ve applied it to the GFS model data. Probably not ready for prime time, but here’s the latest GFS model snow accumulation forecast based on this very experimental algorithm—
Major Snowstorm Increasingly Likely
Posted Friday 02/20/26 @ 5:18 PM —The models have come and they are all moving towards a major snowstorm for the Delaware Valley and NJ late Sunday into Monday. Details about the timing are below.
First, as described earlier, I’ve resurrected my favorite model for snowstorms in Philadelphia. This will likely be its last snow forecast for our region, since the model will be retired sometime this year when the RRFS becomes operational. Here’s the NAM snow forecast—
I’ll get more into the details as the weekend progresses, but I’m going to cut to the chase and post the latest Model Blend Mean and Median foreasts—
Here’s the NBM median forecast. I’ve been leaning towards the median forecast, but I think it may be too low. Somewhere between the median and mean sounds about right.
Major Snowstorm Increasingly Likely
Originally Posted Mon 7:59 PM — This weekend’s weather will start tranquil and will likely end much less so.
Here’s the setup as shown in current water vapor imagery—
