GEOS5ftw

This reminds me a bit of the evolution of the early January 2022 storm. Which the GFS also caught onto a few cycles before EC. Not sure it matters (probably not), but something to consider.

1 hour ago, WxMan1 said:

SIAP...This will be the first 'test' for the growing list of machine learning runs from the ECMWF suite. Verification has been pretty good, especially for some. Not going to get the specifics (i.e. ptype, snow totals, etc.), but at least we can see what the ML runs have in terms of the track into the medium range.

https://charts.ecmwf.int/?facets={"Product type"%3A["Experimental%3A Machine learning models"]}

I mentioned this yesterday...it looks like the 00Z ML suite also shifted significantly S/E with the low pressure compared to 12Z yesterday. Which tells me (from this extremely small sample size) most of the trend is a result of changes in the initial state, not biases in the ECMWF model physics. 

Those 1994 ice storms were what got me into weather which eventually became my career! I do remember 93 but it was 94 that I got into watching TWC all the time.

Two distinct memories from that year (all these are from near IAD) - the MLK day ice storm which started out as snow in the morning, then progressed to sleet and freezing rain during the day. Just when the temp hit 32 the arctic front came through and froze everything solid...the water that had ponded in my backyard literally become a skating rink (I had taken ice skating lessons a couple years earlier, and jammed my feet, which had grown a good deal in those 2 years, into my skates just so I could say I skated in my yard!). This week had a few sub-zero low and single digit highs, and there was a clipper that came through later in the week that dropped 1-2 inches of some of the lightest, fluffiest snow I have ever seen.

Mid-February there was a huge sleet storm - I want to say that the piles of sleet that bounced off the roof created 2 foot "drifts" around the house but the accumulation elsewhere was more like 4-6 inches. 

Then we had 2 inland runner storms, one in early january and another in early march, which were 3-6 inch type deals with wet snow, although I remember hearing that north and west did much better in these.

FWIW, I took a look at the AI models that ECMWF is running here: https://charts.ecmwf.int/?facets={"Product type"%3A["Experimental%3A Machine learning models"]}

With the caveat that they only produce the 850T/500 height/mslp fields, to my eye they all look pretty similar to the Euro op at 12z, if not more NW/warmer. I'm going to try to follow them this winter to see if they are less susceptible to trends than the physics-based suite.

Spire is running a similar model to the GFS ( I think) but with additional soundings from their GNSS-RO satellites. I haven’t seen accuracy scores so it will be interesting to see how it performs this winter.

The AI models (pangu, fourcastnet, etc) are all trained on reanalysis. From my perspective of both seeing a small bit of their output and published verification scores, they can do a bit better than deterministic and similar to ensembles in the 5-10 day range. The real advantage is how fast they run. But they only produce a limited number of output fields and precipitation is a known deficiency.


.

There's a million moving pieces that have to come together for it to work next weekend but the GFS and para are seeing the effects of the block now. Ensembles should have some good hits I would think.

Can confirm, some big aggregates mixed with rain in Wheaton.

39.04, 77.02

A few icy flakes mixing in with the heavy sleet. About 0.3" sleet accumulation so far.

I don't think the high res windows have been mentioned but they are coming into range. I took a look to see if the ptype is correlated to banding structure at all, and there's some evidence for this. In general, ARW and NSSL are warmest aloft (similar to NAM), probably sleet/snow mix or sleet by 12Z for the I66-I70 zone, whereas NMMB and HRRR are colder aloft and mostly snow north of EZF. Interestingly the ARW/NSSL are colder at the sfc than the NMMB and NAM. Not sure how this aligns with typical biases. HREF output should be really useful starting at 0Z tonight.

I'd wait for the 3km which should be in range for most of the storm now. A lot of these subtle mid level temperature changes are related to banding which the 12km NAM does not resolve as well as 3km. Could very well be that heavier precip=snow and lighter precip=sleet in the transition zone.

Just now, Baltimorewx said:

May not be puking fatties but HRRR at end of run has snow for everybody at 11z...earlier start for folks west

Yep, sleet line is way down in southern VA. I think in previous storms this year, the HRRR has been a little generous with snow vs. sleet compared to verification, though.

@Ji UKMet keeps DC below freezing for the entire storm. Main sfc reflection is along the coast instead of the OH valley. First of the more reliable guidance (though UKmet does have some wacky solutions out in time it seems) to really do that, so I'm skeptical. Like you said probably more sleet than snow but 850s don't look too bad. Probably a warm layer above that.

Will be really interesting to see the 12Z RAOB out of IAD. The radar indicated melting layer seems to have collapsed but the lowest few hundred feet still must be torching. A few miles north of 495 here, my PWS is reading 36.7 after starting out near 40 when rain began. I've stepped outside a few times and some very wet flakes are mixed in but still seems to be mostly rain. Seems to be an analog to the early Jan storm last year that gave us 1-2 inches so that's my bar.

eta: 0.15 inches of rain so far in 2.5 hours sinceit began

Para took a step toward the Euro idea IMO. Despite a similar H5 look to the op GFS early on, the CAD is better represented (which should be the case with the para's finer vertical resolution). This feeds back to a stronger, earlier coastal development. Need that H5 to come across a bit further south to really be in business.

eta: The low still escapes east - I was just comparing the H96-108 evolution. Also, seems to be setting up a repeat scenario D7-8.

3 minutes ago, SnowDreamer said:

I've never heard of this technique... Are you using trig and distance from the dish to estimate here? 

Pretty much...I don't know of a site that gives good range rings (COD has 50nm) so it could be off. There is a more formal technique called quasi-vertical profiling where you take the average of the higher elevation scans (e.g. 10 degrees) and plot over time - really useful for seeing how the melting layer evolves over time and other things like dendritic growth. It's becoming more popular in research papers but I don't know of a site that does it in real time.

eta: the obs of rain/drizzle make sense if the layer is at 3000ft- that's probably too shallow below that to refreeze into sleet pellets.

On the dual-pol radar correlation coefficient, there is indication of melting at around 3000 ft (eyeballing the 3.1 degree scan). Something to watch if this disappears as the heavier precip moves in.

GFS showing my concern from last night about the lingering low up in PA preventing the cold press in the boundary layer. Thus we rely on dynamic cooling which is going to be very dependent on where the bands set up (if this gets within meso range, expect to see colder sfc temps line up with heavier precip). Weaker system means we lose this effect and get rain. Good news I guess is that this is the bias I'd expect the GFS to show at this range and the para looks good. I'd still rather see the progression that Canadian was showing last night - it moved the low out of PA faster and had a better CAD setup ahead of the storm. But it is trending towards the GFS setup at 12Z.

I'm not a data assimilation scientist although I interact with that community from time to time. So take this fwiw. 

All of the global centers pretty much ingest the same sfc obs, aircraft obs, satellite obs, and radiosondes. ECMWF has a bit of an advantage because they can use some Chinese and Russian satellites that NOAA is prohibited from using, and I think they have a longer time window for when they allow observations to roll in (this is part of why it comes out later). Where they differ is how they adjust the initial state...they all start from the previous forecast but how much impact observations are allowed to have to change this state, and how observations at one time/location impact the fields at different times/locations depend on the assimilation technique.

The radiosondes have the largest impact (https://gmao.gsfc.nasa.gov/forecasts/systems/fp/obs_impact/), so that lends some credence to giving more weight to the 12z/0z runs if they show a big change, but  there is plenty of other independent info going into the 6z/18z runs.

Biggest concern I have with the GFS is that the primary low from Monday hangs around in PA which prevents any cold from draining into our area. There's really no CAD ahead of the system. So we rely on dynamic cooling as the low bombs out. Canadian looks like it is weaker and faster with the early week system and has more of a classic CAD setup. Big differences apparent at 96 hrs.

Yeah, the para seems to be warm at the start but is slow enough with the southern wave that cold bleeds in and changes most to mix/snow. A little different than the euro front end thump. This sort of evolution isn't common here but to me makes sense with the overall suppressive flow once the NS wave gets out of the way. The whole setup seems pretty sensitive to the relative timing and strength of these 2 features, so I don't think a solution locks in until we are 72 hours out or so.

Just now, WxUSAF said:

5” thump to mix to dryslot? I’d hit that. Interesting that this starts late Monday morning on the euro now. Speeding up.

Yeah, that's the main difference I see on this run - the southern shortwave speeds up and is a bit more neutral than positive tilted, which leads to a strong sfc reflection. The NS didn't change that much, but this progression gives cold a little less time to press in from the N.

I was also questioning the negativity earlier. The pattern doesn't really change until we get high pressure dominating in Canada. That doesn't happen until the northern stream low moves through days 5-6. There's been nothing but low pressure up there since the pattern broke down after the mid-Dec storm. That's a direct result of the Pacific torch but all of the ensembles have been forecasting a complete reversal and hasn't been pushed back in time for several days. That's a direct result of the block and favorable changes in the Pacific as the flow blasting onshore weakens and reverses. That should give every wave a better track and cold air to work with.

Also, I was compiling some climo for MBY from my own records and nearby CoCoRAHS which should represent most of the near-beltway suburbs since 2005. Pretty much every year has featured a 3-6 week stretch without ANY accumulating (>T) snow in mid winter...usually from mid-Dec to mid-Jan, or early-late Jan, or didn't even get going until mid-January. So this isn't a new or unusual pattern. Maybe different if you go further back but it's typical the past 15 years, not just since 2016.

4 minutes ago, HighStakes said:

I was thinking about that 2 week period quite a bit leading in to this winter. maybe it's just a hunch or wishful thinking but it is on the table. Our good window really doesn't start until the 15th. This upcoming week still may workout but it looks like we got a bit of a head fake. When our official window gets here its look like well have 2-3 weeks to score. Hopefully there is a reload and we get extend into first 2 weeks of February. We know it has to break down but hopefully we max out on length of favorable indices. Whether or not we enter a 2-3  week stint of deep winter like in 2000 is the million dollar question.

For those too young to remember I'll give recap of how that period in January 2000 played out. Horrible 1st 2 weeks of January. Pattern changed on the 18th. Very cold cloudy day with a small clipper that dropped only a half inch of snow but snarled traffic around DC. Only virga from Baltimore northward. Then a nice quick moving system dropped 3-6 inches to kick start the run. The next day was very windy and cold. There was a potential storm on the 23rd but it failed and resulted in flurries. Then of course the 25th we all know about. The next storm was on the 30 th superbowl sunday. Forecast busted in a positive way for northemwest burbs. 4-8 inches fell and never changed over even though most forecast were for 1-3 and a change to a mix then rain. Stayed chilly after that storm for a few days with some snow showers then winter ended. I got 26 inches in Reisterstown all in a 10 day  stretch. 

Nice recollection. That first system (Jan 20?) is overlooked because of what came after, but a very underrated storm IMO. It was a pure NS wave that dropped far enough south to work for us. Then it blew up into a huge bomb offshore (like 950mb) and hit Nova Scotia pretty hard, but missed NE. That's why it was so windy afterwards, some of the most impressive blowing and drifting I have seen in this area especially considering the storm total was only 3-6 as you mention. I wonder if the look being advertised day10+ could lead to this type of opportunity.

47 minutes ago, cbmclean said:

That's actually the part of meteorology that really interests me.  I find cyclogensis, precipitation, winds speeds, only mildly interesting, but I am fascinated by temperature.  Unfortunately there just isn't a whole lot of easily available information out there beyond the simplistic guidelines we have just been discussing.  I need to find a text book on "Atmospheric temperature patterns and 500 MB analysis".

Since I don't think it has been answered yet, the 500mb height is a function of the surface pressure and thickness (which is proportional to temperature averaged over the layer). So, BN heights can be the result of BN temperatures from the sfc to 500mb, BN surface pressure, or some combination. When the 500mb height is BN but sfc is AN, that tells me the sfc pressures are low and/or the sfc temp is higher than would be expected for a given thickness. This makes sense since Pacific maritime air masses have steeper lapse rates (warm sfc, cold aloft) than continental/arctic origin air (cold sfc, cold aloft). Think about it - air coming from the Pacific more or less assumes the SST after many days. There isn't enough time when these travel over North America to cool radiatively at the sfc. Also, the latent heat added from orographic precip over the Rockies actually results in a warming of these air masses (chinook effect). Contrast that with Arctic air masses - the cooling is strongest at the surface (surface radiates heat much more effectively than atmosphere), so these tend to have a strong inversion. Further, when you get an EPO ridge the cold air doesn't have to cross the Rockies, it comes down the eastern slope from the north so no latent heat gets added (not that there would be much considering how dry these air masses are).

Models are really good at moving air around - I'm not sure how well they represent the radiative heating/cooling though. It's pretty complex and depends on the surface properties, water vapor, etc. I remember reading a paper about the formation of deep cold air masses, and the authors found that you get "diamond dust" ice crystals - basically very light snow w/o clouds - forming below -40C (or something really cold), and these radiate heat more effectively than the air itself. At really long lead times if these processes are not represented properly it could result in a odd looking temperature for a given flow pattern. There are modeling experts on this forum who know way more than me (I'm from more of a radar/remote sensing background) , so I'm happy to be corrected.