Trom

Interesting height rises to the NE late in the GFS run (day 15 posted) Could this finally be the impact of the SSW hitting the trop? Will need to see if this sustains in future runs and whether it gains any cross model support once it comes into the ECM 10 day timeframe. Thought it was worth posting given the reasonably bullish Met 6+ day forecast

My firm has an office in LaCrosse Wisconsin. There's not a lot going on in most of Wisconsin other than corn and dairy. The temps get so low that it rules out being outside for any period of time as any exposed skin starts to freeze. I have been ice fishing on a frozen Mississippi. Massive cold period about to hit the mid-West in the next few days. LaCrosse will be -20C in the day and closer to -30 at night by the weekend.

Yes, the operational is very close to the mean for 7 days out. Recent runs the op and ensemble means haven't been. Does this herald some consistency and perhaps less volatility on the ECM in future runs? Time will tell.

I think it must be ensemble members from the GEFS

I agree about the no panic it's just that I'd rather see a cold op supported by the ensembles rather than an operational that's at variance with it's ensembles. Hopefully, you are correct and the ECM op moves into alignment with the ensembles. My unease is also compounded by the fact that it's the ECM op that is so out of kilter with its ensembles and GFS output. Another way of viewing it that the volatility caused by Trop/Strat interplay is having a major effect on the reliability of the model output and that higher definition compounds this uncertainty. We did see all models struggle significantly with the quick trop impact last year. This year's been a more normal SSW impact (i.e. slower) but we know it's impacting shortly. Fingers crossed.

Well partly and the fact that the initial input data is tweaked for the ensemble members (control excluded).

I don't think you can ignore operating runs given they are high definition compared to the ensembles and controls low definition. The question is, especially on the ECM, what is the operational picking up on that the lower definition control isn't that causes the direction to be so different.

I always think it's like a break in snooker. You know where the balls should roll but in essence, every break is different. The physics look simple but the outcome is surprisingly variable in snooker. The split SSW is like the white ball and the voretex the reds. You know the vortex will dissipate but exactly where and how remains an enigma.

There were some interesting posts the other night stating that the GFS and parallel are currently experiencing difficulties and low verification stats. The argument was that the Federal shut down in the US means the models are running in auto mode with no intervention. The way automatic data is recorded has changed and due to the shut down the US models have not been adapted to this new system. So issues with initialisation data. This is compounded by the facts that the models need to be manually adjusted for known bias which isn’t happening right now Work has stopped on the GFS parallel. As Whitefox said the parallel is in a testing phase. This is where the model outputs are correlated against actual outcomes and the differences analysed (Root Mean Squared Error ‘RMSE’ for those of you who like regression models). The algorithms are then tweeked over the parallel period to reduce the RMSE and hence increase predictive accuracy. With the Fed shut down this is not happening. For those not hugely into stats RMSE is the standard deviation of the difference between forecast and actual outcome. Low indicating differences a small and a high RMSE indicating big dispersion between forecast and actual outcomes. The upshot is a delay to the parallel taking the place of the existing model and lower verification stats for both GFS and parallel in the meantime. This came from workers at the agency and there was a good article in the Washington Post https://www.google.co.uk/amp/s/www.washingtonpost.com/amphtml/weather/2019/01/07/national-weather-service-is-open-your-forecast-is-worse-because-shutdown/

Well never given they are incorrect at T+0! Good use of perturbations is: 1. Look for how many support the operational run. 2. Look for clusters indicating possible trends. Individual perturbations not hugely useful as they are low resolution and start off with tweaked initial data. So I'm with you Froze. However, pointing that one out I think is more for the holey mother imagine if that came off point rather than anything else.

It's worth noting that the models could not cope with the SSW last year (wave 2). I seem to remember that we had a huge period of volatility in the NWP until after the event. Then the change in the models, the week after the warming, was abrupt. It wasn't something we saw in FI that got closer and closer. So they really struggled with last years wave 2. If this pattern repeats this year then anything showing now for the post-warming period should be taken with a pinch of salt. The models appeared to be reasonable with the forecasting of the warming of the strat but all at sea with regards to the tropospheric impacts. The question right now is are we likely to see wave 1 or wave 2 warming. Wave 2 with it's sheering rather than pushing effect on the vortex gives us a higher probability of cold weather. I would also presume that the models find it harder to model the impact of wave 2 rather than wave 1 due to the more complicated variables of a split vortex (although that is just a hunch). The key variables for us right now are wave 1 versus wave 2 and the likely speed of tropospheric response.

I seem to remember that last year the NWP struggled with the wave 2 event. I seem to recall it was significantly after the event that the models picked up the implication for the troposphere but when they did the impact was marked. Before the event and immediately afterward they did not cope at all well if my memory serves me well. I'm not sure how well they cope with a wave 1 event. Do they find a displacement easier to model than a total split? Anyway, I'd say patience is a virtue right now.

Hmm, I'm not so sure I get sucked into this SSW in January theory just yet. Obviously, the Strat is easier to model as it has less volatility than the troposphere but we are talking a month's time. I'm not so sure that I've seen anything particularly accurate at that range. If true then with a 2-3 week down welling you are looking at late Jan/early Feb period of cold potentially. What is encouraging is that all the main models have pointed to wave one events. I always think of wave one as a nudge to the vortex, compared to the scattering effect of a wave 2. For some reason, I always think of snooker players breaking off. My main point was that it looks interesting in the long term and to remind people that December snow at low level is an anomaly (2010 skewed peoples expectations).

I think that's a result of the two winters we had back to back when the Atlantic just sent low after low in our direction. Must have been 3-4 years back now. Also the cold period in Dec 2010 also has skewed peoples expectations of cold in late Nov/Dec. It's just behavioral bias. Also there's a lot of cognitive dissonance where models that support a poster opinions are highly rated but those that don't are sent to the bin. I may have mentioned, in past years, that I'm a finance lecture and use the contrary post we get on this forum as we approach winter to illustrate behavioral bias. If you are new to the field of behavioral science then the BBC had a good introductory Horizon episode

I agree the low heights to the SW stop the high above us sinking but surely the strat warming in 4 days won’t be quick enough to effect the trop in the timeframe needed to impact this cold snap. That would be an extremely quick response. Interested if you think it could be that quick.

Yes odd - clearly that chart isn't -4 to +4 standard deviations. +/- 2 captures 95% of all observations. So >4 standard deviations would be so infrequent that this can't be what he means. Am I missing something here? Googled +/- 4 sigma and it's 99.993% of observations.

Clearly convective snow can’t be forecast much in advance this spell has underlined. I kind of new that but still have been surprised about the frequency of snow showers over a number of days. Today was leaving South Croydon at 6am to catch a flight to Amsterdam. Amazing windchill here. I also got a picture of a frozen canal but can’t seem to upload it. Back to Gatwick midday tomorrow. Flight currently saying no delays expected. We shall see.

The slower retrogression of the high pressure is having the impact of keeping that low further South.

Movement in the ICON and GFS to keep the low further South on Friday. Got to hope that movement is sustained on future runs as it would keep our area in much colder uppers.

The operating run has greater resolution. The control uses the same initial data as the op but is run in low resolution. The ensembles are run in low definition but have tweeks to the starting data. Purpose of the ensembles is to add support to the results of the operating and control runs. The ensembles can also be used to identify clusters which give us an idea of future trends. What they can't really be used for is analysis of individual perbs with some kind of hope that they will materialise.

2 standard deviations captures 95% of all observations around a mean. Does, of course, depend on the mean being accurate in the first place.

Well GFS 18Z showing a couple of days of South Easterlies which keeps the snow risk this weekend. Then the Atlantic breaks through and giving us a westerly regime. By T72 high pressure is edging in leaving us in a southwesterly regime and hence mild but dry especially to the south of the country. Temperatures up to low/mid-teens for most of England by T90. By T117 were again in a more westerly regime as the high pressure ebbs east. We then seem to alternate between westerlies and south westerlies. Scotland flirts with a short-lived North Westerly by T138 bringing temperatures there down and the rest of the UK is in a North Westerly by T156 which will, therefore, bring a dip in temperatures to lower single digits but nothing that will bring snow away from mountainous areas in the north-west. By T183 it's back to south westerlies. Doesn't look inspiring.

Met 10 dayer

I've got to agree. Looking at it the surface temperatures look too high on Monday for most areas (seems to be around 5c). At the moment to me, it really looks like elevation is needed, even in the North. Snow chances do look to increase as the week progresses but with the flows direction, it all still remains highly marginal away from high ground. As for the south I really can't see anything other than sleet next week. I do hope I'm wrong and those 850's dip a little further to get the right side of marginal for more widespread areas.

Whilst it's relatively quiet on here it's interesting that from a behavioural aspect that the December of 2010 has certainly skewed perceptions that late November can be cold and early December are particularly snowy months but March 2013 has not skewed perceptions that March is particularly wintery. Of course January and February are the historically most wintery for the UK. I wonder if psychologically it's because November and December are the start of winter whilst March is the end (obviously not meteorological seasons before some one calls me out). Anyway letting past events impact on future unconditional probabilities aka the gamblers fallacy is a repeating feature of the models thread come winter time and has been for many a year. So it looks Atlantic driven for the next week, albeit a relatively benign Westerly pattern. Unsettled would be the definition especially the further West you are. Then it's wait and see wether the SSW can shake things up in terms of a last cold shot.