Snow Potential Dec 26-27

[steve392]

Had heavy sleet turn back to big flake snow then back to sleet then snow again.  Up in Dumont, nj and it's plain snow. 

[The 4 Seasons]

Looks like OKX/PHI cut back on amounts and downgraded warnings to advisories for Somerset and Middlesex. Saw that coming, glad we went conservative there

[steve392]

Heavy heavy snows in Dumont, nj right now.  Really accumulating.  Salting isn't going to do anything now. 

[DDweatherman]

On 12/25/2025 at 4:33 PM, PhiEaglesfan712 said:

That has no correlation. NYC cashing in on a snowstorm doesn't mean everyone is going to win come January. For all we know, DC could end up with a final season snow total of 3-4" (see below).

That season had one of the oddest snow distributions ever. Places in upstate NY finished the season with over 150 inches of snow, while places like Baltimore and DC finished with less than 4 inches of snow.

I’m late on my reply. I meant that in no meteorological way. It was really just supporting my northern brethren in hopes that we get a miller A or a big hybrid low that blesses the masses later this winter!

[TriPol]

Winter Storm Warnings are canceled for the immediate Metro area. 

[donsutherland1]

17 hours ago, donsutherland1 said:

I'll do the 18z HRRR, too. I think that model is often cold-biased, but let's see how it does.

18z HRRR.

I will also compare the 18z NAM and 18z HRRR.

My point estimates:

Bridgeport: 8.0"
Islip: 7.9"
New York City (Central Park): 5.8"
Newark: 4.8"

18z NAM vs. 18z HRRR: 

Here's how I fared against the two models:

Final Accumulations:

Bridgeport: 7.1"
Islip: 6.6"
New York City-Central Park: 4.3"
New York City-JFK Airport: 4.1"
New York City-LaGuardia Airport: 4.1"
Newark: 4.2"

The low end of the 4"-8" idea worked out. The loss of precipitation from the weak lift that developed in the spacing between the surface and mid-level low likely deprived the region of several inches of snow. It also allowed for some intrusion of sleet into the greater New York City area for a time until the lift improved.

[EastonSN+]

9 minutes ago, donsutherland1 said:

18z NAM vs. 18z HRRR: 

Here's how I fared against the two models:

Final Accumulations:

Bridgeport: 7.1"
Islip: 6.6"
New York City-Central Park: 4.3"
New York City-JFK Airport: 4.1"
New York City-LaGuardia Airport: 4.1"
Newark: 4.2"

The low end of the 4"-8" idea worked out. The loss of precipitation from the weak lift that developed in the spacing between the surface and mid-level lows likely deprived the region of several inches of snow. It also allowed for some intrusion of sleet into the greater New York City area for a time until the lift improved.

You nailed this one Don. Great job!

[BxEngine]

24 minutes ago, donsutherland1 said:

18z NAM vs. 18z HRRR: 

Here's how I fared against the two models:

Final Accumulations:

Bridgeport: 7.1"
Islip: 6.6"
New York City-Central Park: 4.3"
New York City-JFK Airport: 4.1"
New York City-LaGuardia Airport: 4.1"
Newark: 4.2"

The low end of the 4"-8" idea worked out. The loss of precipitation from the weak lift that developed in the spacing between the surface and mid-level lows likely deprived the region of several inches of snow. It also allowed for some intrusion of sleet into the greater New York City area for a time until the lift improved.

This is kind of awesome to see, thank you for doing that work!

[SnoSki14]

So the no 4" streak is officially over then 

[donsutherland1]

Just now, SnoSki14 said:

So the no 4" streak is officially over then 

For snowstorms, it's over. Finally. For daily snowfalls, it, unfortunately, continues.

[donsutherland1]

As forecasting accuracy requires, among other things, an objective assessment of the guidance, I've expanded the comparison of the 12/26 18z NAM and 12/26 18z HRRR. The major point of contention had been whether the NAM was providing a realistic forecast for the New York City area. It did not fare well.

But what over a larger area? Again, the HRRR outperformed the NAM, but its forecasting edge was less decisive. The HRRR was better in assessing where the 4" or above snows would fall from New York City southward. The NAM confined such snows to the north and east of New York City. However, the HRRR was too aggressive in pushing 2"+ snows past Allentown and into Philadelphia. The HRRR also did better in such locations as Albany and Boston.

12/26 18z NAM:

12/26 18z HRRR:

Outcomes for 12 Cities:

[qg_omega]

23 hours ago, donsutherland1 said:

For fun, let's compare the 18z NAM with the NWS for select locations that provide snowfall reports:

18z NAM:

NWS Forecast:

I will be looking at the other 18z guidance. For now, I still think a general 4"-8" snowfall is likely in and around New York City (including all of the above locations).

Verification scores?

[donsutherland1]

31 minutes ago, qg_omega said:

Verification scores?

Against the NWS? I will run them when I get back in this evening. This wasn’t the NWS’s finest forecast.

[eduggs]

12 minutes ago, donsutherland1 said:

Against the NWS? I will run them when I get back in this evening. This wasn’t the NWS’s finest forecast.

Couple minor comments: on the 18z NAM snowfall map I looked at on pivotal at 10:1, Islip was 5.0, NYC was 2.6, and Newark was 0.8. But the colored graphics look wonky - I'm not sure how accurate they are. Point and click shows about 2" for Newark. It shouldn't be much different than MMU considering their relative lat lon positions.

Secondly, I don't believe those numbers count sleet. I believe the algorithm ignores accumulated sleet and only displays the snow equivalent. So some sleet accumulation (NYC and EWR) would lead to underestimation.

Since snow accumulation is a very inexact metric to both forecast and measure, I don't believe a subset of reported snow totals is a good method to assess model verification.

[eduggs]

Middletown is a funny case. For the Tue event the NWS hoisted a last minute WAA and the area ended up with warning snows. For this relatively long lead WSW event, I believe that area ended up in the upper tier of advisory snows. Snow is very difficult to accurately predict.

[donsutherland1]

2 hours ago, qg_omega said:

Verification scores?

Here you go:

[donsutherland1]

1 hour ago, eduggs said:

Couple minor comments: on the 18z NAM snowfall map I looked at on pivotal at 10:1, Islip was 5.0, NYC was 2.6, and Newark was 0.8. But the colored graphics look wonky - I'm not sure how accurate they are. Point and click shows about 2" for Newark. It shouldn't be much different than MMU considering their relative lat lon positions.

Secondly, I don't believe those numbers count sleet. I believe the algorithm ignores accumulated sleet and only displays the snow equivalent. So some sleet accumulation (NYC and EWR) would lead to underestimation.

Since snow accumulation is a very inexact metric to both forecast and measure, I don't believe a subset of reported snow totals is a good method to assess model verification.

On Pivotal, the numbers don't count sleet. On Tropical Tidbits, they do, but sleet is treated the same as snowfall, which inflates the totals. Both approaches lead to inaccuracy, since snowfall figures are the sum of snow and sleet.

[eduggs]

29 minutes ago, donsutherland1 said:

Here you go:

Where did you get your 18z NAM numbers?

[eduggs]

The numbers from the 18z 12/26 NAM run on Pivotal (which doesn't add anything to the totals for sleet) were:

Bridgeport: 5.8
Islip Airport: 5.0
Central Park: 2.8
Newark Airport: 0.8

However, the annotated totals shown for NWS reporting stations don't match the colored boundaries drawn on the map. I've marked Newark Airport and MacArthur Airport with red dots as an example. Based on the graphical boundaries and the legend, the estimated NAM totals are:

Bridgeport: 5.5
Islip Airport: 5.5
Central Park: 2.8
Newark Airport: 1.8

So now I'm wondering how the values shown on Pivotal get populated. Is the data drawn from the exact lat lon of the reporting station cross referenced to the raw NAM data, or is there some kind of apprimation going on...?

 

[psv88]

51 minutes ago, eduggs said:

The numbers from the 18z 12/26 NAM run on Pivotal (which doesn't add anything to the totals for sleet) were:

Bridgeport: 5.8
Islip Airport: 5.0
Central Park: 2.8
Newark Airport: 0.8

However, the annotated totals shown for NWS reporting stations don't match the colored boundaries drawn on the map. I've marked Newark Airport and MacArthur Airport with red dots as an example. Based on the graphical boundaries and the legend, the estimated NAM totals are:

Bridgeport: 5.5
Islip Airport: 5.5
Central Park: 2.8
Newark Airport: 1.8

So now I'm wondering how the values shown on Pivotal get populated. Is the data drawn from the exact lat lon of the reporting station cross referenced to the raw NAM data, or is there some kind of apprimation going on...?

 

You can’t go by the maps, use the lat/lon

[eduggs]

1 hour ago, psv88 said:

You can’t go by the maps, use the lat/lon

But that's the question. How does Pivotal retrieve the raw data? It's suspicious that the NAM would output so much more "snow" for MMU than EWR considering their relative locations and the general snowfall distribution it was predicting. I'm wondering if it pulls the raw data from a location further southwest than the airport.

[eduggs]

Don was ripping the NAM before and during yesterday's event. It should be obvious to anyone that the NAM (from ~60 hours out) performed better than the NWS and most other guidance for our area. Don was wrong. If he wants to do a verification of the NAM, he should do it with a more precise parameter than reported snow accumulation. At the very least he should use more than 4 stations. I think the NAM is getting hosed in Don's "verification." A few of the entries look wrong. There's also the issue of sleet: EWR and possibly Central Park experienced some sleet yesterday evening. If this was the dominant ptype during any time point, pivotal will record it as 0 snowfall for that period. But since sleet accumulates, using the pivotal algorithm will under-predict accumulated snowfall.

I think the entire purpose of the verification when it was originally proposed was to bash the NAM, assuming it would fail. But it did not. The writing was on the wall early yesterday (HRRR, RAP, ECM, radar). Some noticed. Others doubled down .

[psv88]

3 minutes ago, eduggs said:

Don was ripping the NAM before and during yesterday's event. It should be obvious to anyone that the NAM (from ~60 hours out) performed better than the NWS and most other guidance for our area. Don was wrong. If he wants to do a verification of the NAM, he should do it with a more precise parameter than reported snow accumulation. At the very least he should use more than 4 stations. I think the NAM is getting hosed in Don's "verification." A few of the entries look wrong. There's also the issue of sleet: EWR and possibly Central Park experienced some sleet yesterday evening. If this was the dominant ptype during any time point, pivotal will record it as 0 snowfall for that period. But since sleet accumulates, using the pivotal algorithm will under-predict accumulated snowfall.

I think the entire purpose of the verification when it was originally proposed was to bash the NAM, assuming it would fail. But it did not. The writing was on the wall early yesterday (HRRR, RAP, ECM, radar). Some noticed. Others doubled down .

You are going to die on this hill aren’t you? Lol. Everyone has their thing I guess. 95% of us enjoy the snow. The other 5% are never happy and find ways to complain. You seem to be in the 5%. 

[eduggs]

Just now, psv88 said:

You are going to die on this hill aren’t you? Lol. Everyone has their thing I guess. 95% of us enjoy the snow. The other 5% are never happy and find ways to complain. You seem to be in the 5%. 

I feel like you don't actually read what I write, which is a shame. Being an "old-school" poster, I put a lot of thought into what I write and it's often more than 280 characters. I thoroughly enjoyed the snow yesterday evening in Putnam County and then late last night in Morris County. I'm certainly not complaining. 

I'm trying to make a point about the NAM so that people are less likely to make the same mistake during the next event.

[donsutherland1]

29 minutes ago, eduggs said:

But that's the question. How does Pivotal retrieve the raw data? It's suspicious that the NAM would output so much more "snow" for MMU than EWR considering their relative locations and the general snowfall distribution it was predicting. I'm wondering if it pulls the raw data from a location further southwest than the airport.

I also used Bufkit. Its numbers for EWR were absurdly low. One can get a good picture of the snow growth and insight into ratios from its soundings.

The focus for me was the NYC region given the subforum involved. The NAM did not do well in this subforum. I don’t always disregard it, but in this case I did. It was an outlier even against the vast majority of individual EPS and GEFS ensemble members. That’s what led me to dismiss it. It is quite rare where the ensembles don’t capture the cluster of reasonable solutions, one of which ultimately verifies.

The GFS was overdone. For several days, it was clear that the NYC area was in line for a 3”-6”/4”-8” snowfall regardless of the earlier RRFS A runs and the NAM’s solutions. The only real drama was for areas south and west of NYC and areas in parts of New England. 

[eduggs]

1 hour ago, donsutherland1 said:

The focus for me was the NYC region given the subforum involved. The NAM did not do well in this subforum. 

What are you basing this on? Can you share your data? My assessment does not match yours at all.

I thought the NAM did great. The average of its last few runs before go time was more accurate than any other model IMO. It picked up on the track of the 700mb low and delaying its weakening. It correctly depicted the resulting dryslot across the southern tier of NY. It brought sleet past the NJ-NY border (correct). It was the first model to target and then consistently target the NJ-NY border as the dividing line between minor and significant snow, particularly with the initial overrunning. It (along with the RDPS) correctly highlighted the low-level lingering snowfall into Saturday that the globals undermodeled. Just overlaying its QPF forecast and clown maps with the reported snowfall matches up much better than everything else.

The GFS, GEFS, CMC, RDPS, GEPS, HRRR, UK, ICON, and ICON-EPS were all consistently too far southwest with the heaviest precipitation, total snow, and all-snow zone. Even the ECM had too much snow in most of NJ until the very end. It also failed to show the gradient across Long Island.

Hopefully you are relying on more than a single NAM run and a wonky bufkit output for one station.