Jump to content


  • Content Count

  • Joined

  • Last visited

Community Reputation

243 Excellent

Profile Information

  • Gender
  • Location
    South Croydon

Recent Profile Visitors

2,277 profile views
  1. 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.
  2. 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/
  3. 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.
  4. 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.
  5. 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.
  6. 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).
  7. 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
  8. 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.
  9. 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.
  10. 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.
  11. The slower retrogression of the high pressure is having the impact of keeping that low further South.
  12. 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.
  13. 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.
  14. 2 standard deviations captures 95% of all observations around a mean. Does, of course, depend on the mean being accurate in the first place.
  15. 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.