MegaMike

00z HREF. Sum the two to get your total:

Likely the SREFs, but the NBM consists of 80+ different, individual ensemble members. I'm going off memory here, but it includes 10 SREF, 30 GEFS, 50 EPS, 3kmNAM, GFS, RAP, the high-res ARW models, and maybe 1 or two different (experimental) diagnostic models. The models incorporated into the NBM are lagged by a couple cycles too so it's not ingesting the most recent simulations. Honestly, I get why people get so confused with it.

Ensembles are usually more truncated (courser, less dz layers, greater time step), but a model (NAM, GFS, etc...) is just a name given to a set of model configurations. The core (dictates how the atmosphere acts) is one aspect of multiple modeling options/schemes. Here's a table of some NCEP modeling configurations: https://www.emc.ncep.noaa.gov/users/meg/home/table.html

That's interesting. Would you oppose retiring the SREF? Looking at the evaluation webpage, the GEFS currently outperforms the SREF.

I hate to call any model useless, but if my name was 'Big Balls' and I worked at DOGE, I'd probably get rid of the SREF (and maybe the CFS in favor of AI). It's consistently overamped during costal storms and I never use it either. It's configuration isn't ideal for severe weather/convection too (16km dxdy/only 40 vertical levels). The one + about it, the individual members are very diversified. Some fun combinations of schemes which promotes a huge model spread (ideal for an ensemble!)... It'd be replaced by the RRFS though. It's quite fun to work with! Unfortunately, I do think WRF will be replaced by something like the MPAS. In terms of efficiency, the MPAS is better (takes less resources to run)... No need to feed bcs onto 3 different grids. It's the future and we're already experiencing it (FV3 core; ignoring AI). Generally, we're told not to configure modeling systems with a resolution ~12km. It's a gray zone for convective parameterization... Too fine for parameterization, but too course to explicitly predict it. We need NAM 12km to feed NAM 3km bcs. It's a sacrificial lamb. <4km should do fine with convection.

lol I get it. Yea, I remember another meteorologist mentioning that they didn't like the WRF-NMM/ARW models. I've worked with them for a decade, so I guess I'm a bit biased. I think @vortex95 works with the MPAS(?) so I wonder what his thoughts are regarding the WRF models. The RSM must be before my time. I'm not familiar with it

The table comes from NCEP. I just formatted them differently so I can sort the weights. My opinions: They need a modeling system to encompass the entire country, so the HRDPS wouldn't be applicable. I think the ECMWF is included as one of the 50 EPS members. (a) Too soon for the AI models as they recently became operational. (b) I'm guessing they didn't include the RGEM because of a licensing/boundary issue (cells close the boundaries can produce wacky results). (c) The ECMWF outperform the UKMET so no need to include it. (a+b+c) More importantly, everything must be post-processed onto a constant grid - it's not worth including a single diagnostic model when the ensemble itself already composes of 80 members. You'd just be complicating things with little in return if you include modeling systems with different dxdy configurations.

For 20-42hrs? I agree that they can get pretty erratic past one daytime cycle and I've mentioned that before. It's a weird gray zone for modeling (24-48hrs), but I'm sure NCEP created those weights with solid reasoning/evidence/support.

Mesos should definitely be weighted more as you get closer to an event... Small scale features impacts sensible weather and courser models cannot resolve them/assimilate them well. A lot of members repeat this tidbit each storm, but the s need their dopamine fix. Here's how NCEP weighs the models for the NBM (snowfall specifically because I'm obsessed with it). Regardless, the weights won't vary much by variable. Since we're ~20-42hrs away from the event, I sorted that column by importance (values are in %). Note: The top 5 model weights are all mesoscale modeling systems and the GFS (diagnostic) is only weighted by 3%.

I could see him going to either Providence, Boston, central NJ, or somewhere along the Cape. I bet they announce the location later tonight. 18z HRRR gives Hartford higher percentages of lightning than Providence... 3-6% from 12am-3am on Monday. NYC too at 3-7% from 11pm-2am Sunday-Monday.

So impressive! The 18z HRRR has a 3-5% chance of lightning (throughout the entire atmosphere) for Providence from 12am-3am on Monday.

Shameless bump. I ran an analysis simulation w/WRF for the 78' blizzard that evaluated well

HREF and NBM be honking. Still not completely in range of the HREF though (still snowing):

If I had to take a wild guess, the SREFs are likely beefing up the mean.

It's so hawt' I can barely contain myself, Tippy!

I'm excited to test my weather application for this event. Here's what it looks like for Boston (NBM): And the HRRR: I stole the icons from The Weather Channel

13z NBM is pretty wild. The mean(!!!) has a large area of 24-30" across the Boston/Providence area (even 30-36"). The diagnostic ratios are likely too high considering winds will be strong. EDIT: for NYC weenies, this is 48hr-snowfall accumulation maxed out for SNE... The totals are higher for the NYC area, but you need an earlier image to capture it better.

Man. I was thinking whiff/graze yesterday at noon. It's wild how well this has trended since yesterday. The GFS looks so hawt'

Still ongoing, friend: While on the subject, I wrote to Don and someone else about how snowfall is determined by the NBM several weeks ago. For clarity, I mentioned that the weights (available here: https://vlab.noaa.gov/documents/6609493/32850490/CONUS_SNOICEACCUM.pdf) temporally vary. On top of time-varying weights, the ensemble members incorporated into the mean are lagged by ~1-2 cycles (depends on modeling system)... To see which models/cycles are included by the NBM at any forecast time, you can use this link: https://blend.mdl.nws.noaa.gov/nbm-dashboard Additionally, it takes ~1 cycle to post-process the NBM. Therefore, the most up-to-date NBM snowfall product will have a lag of ~2-3 cycles. This is where the confusion is imo.

I never imagined the GFS/GEFS scoring a coup against almost every available modeling system. Yet, here we are... As we continue to see westward trends, the NBM (13z) is starting to beef up mean-snowfall considerably. This doesn't include today's snow event (map goes from Sunday 1am to Tues. 1am). As modeled, it looks like it'll start snowing b/n ~1pm-7pm on Sunday and ending late Monday night across SNE. For clarity, the NWP members incorporated into the NBM are (generally) lagged by 1-2 cycles... For example, the cycles utilized for the 13z NBM are 00z (EPS), 06z (GEFS), and 03z (SREF). Additionally, on DESI, the 13z NBM is the most up-to-date cycle available, so on top of the 1-2 cycle delay, you have another 1 cycle delay for post-processing... I understand why people are so confused with the NBM's snowfall depiction (not even considering how weights are incorporated which vary by time). You can see which cycles are used at https://blend.mdl.nws.noaa.gov/nbm-dashboard. In the past, I only added the weights (https://vlab.noaa.gov/documents/6609493/32850490/CONUS_SNOICEACCUM.pdf). Based on the most recent model cycles, the mean will bump up next cycle too.

Yea, unless the ECMWF/EPS improve at 12z, I think this will be a graze, at best. Definitely tossing the GFS/GEFS.

At this point, regardless of trends, I'm leaning towards the ECMWF/UKMET and their ensembles (graze/OTS)... Especially since they're both near misses at the moment and the CMC backtracked a little bit. No matter what I analyze in terms of evaluations (fields, isobaric surface, space, and time), the GFS performs comparably to the CMC and JMA. The ECMWF and UKMET are 1 and 2, respectively. Maybe the UKMET went to rehab over the past few decades and became an accomplished businessman?

I agree with you. Ignoring trends, this looks like a graze verbatim. Most ensemble means (I think Ryan posted probabilities earlier) barely, if at all, get the 24hr-QPF 0.5" contour to the Cape/Islands. Definitely keep expectations low, for now.

Agreed with dendy and Coastal. Better than the NAM, but would kick OTS. I would say it looks worse than its 06z cycle too.

It's still ~5 days out. Still plenty of time for trends, as Mr. Sey-Mour alluded to. Would ease my if the EPS improves.