Trom
-
Posts
226 -
Joined
-
Last visited
Content Type
Forums
Blogs
Gallery
Events
Learn About Weather and Meteorology
Community guides
Posts posted by Trom
-
-
1 minute ago, sebastiaan1973 said:
Well. Perhaps this was in the past so, but there seems to me enough vertical layers in e.g. EC-oper, to know what's going on. It's just a part of the atmosphere.
Yes but the algorithms have relatively little training data on these events. So in my opinion the jury's out with regards to the effects. You could be right and ECM has it nailed but the models certainly did not in 2018 until sometime post event. I'm not sure the underlying algorithms can have improved that much since then given the lack of events. I think what is true that with the increased resolution in the vertical layers we will learn more from this event going forward which will improve modelling going forward.
-
1
-
-
14 minutes ago, bobbydog said:
I do chuckle when I see comments like "the models haven't got a clue" or "they haven't factored in the SSW yet".
They do and they have....
They are computer algorithms fed with a vast amount of information regarding current atmospheric conditions. The fact they vary so wildly right now, demonstrates the fact that these conditions are very volatile due to the fact we have an SSW happening at this moment and subject to change at very short notice, hence the multitude of possible outcomes we are seeing. They are also fed this information at different times. So what data one model is given, may have changed by the time another is given its input.
So, what the models are showing us is that the weather itself is currently highly unpredictable....
Whatever happens, just watch and enjoy the show....
Yes but how the algorithms cope with such rare event's it uncertain. It's way away from climatic norms.
-
3
-
-
1 hour ago, John88B said:
Serious question from a complete novice when it comes to reading the models but is it worth even looking at the models for a period of time if a SSW can shuffle the deck so much so to speak? Thanks in advance.
Yes and no. My experience has been that initially after the SSW (which still is 2 days out), the models struggle to process how the downwelling will impact the trop. So you get a lot of volatility between both models and runs. Then the SSW will take between 2 week and 2 months for the impact to be felt. In 2018 we had a QTR (quick trop response) and we actually felt that impact within 2 weeks. The models started to show some consistency again fairly quickly (post event). I'm not a strat expert but it's been suggested that the response from strat to trop will be quick this time too. So I'd expect a lot of volatility (people on here like to refer to it as Shannon entropy but don't get me going on that one), between runs and between models. If the response is quick expect to see the volatility dropping over the next week. Posters on here correctly state that the models know there is going to be a warming but the main issue is the interaction of the strat and trop and the outcome. We know reversal of wind is likely as is blocking but exactly where the blocking sets up seems to take longer to resolve. I think the key to remember is no poster or model has this dialled in. SSW have only been studied in earnest fairly recently.
So don't despair if runs flop from ultra cold to mild. If we are lucky with the SSW (and we seem to be around 2/3rds of the time), then expect to see colder solutions impacting the models some time post 5th of Jan (speed of the propagation being key).
There are some great posters on here so try to follow them. It takes a little bit of time to understand how to read a model but once you can you get a better idea of who to pay credence to.
-
4
-
1
-
-
Just now, Day_9 said:
Cheers, where can we easily reference verification rates?
Verification stats are for the entire northern hemisphere. What they won't tell you is how a model copes with the additional variables caused by a SSW event. There's just not a huge amount of data on it. For example the last one in 2018 caused the models a lot of problems and it took an number of days to settle down. It's also worth bearing in mind that the models change overtime in terms of resolution and the base algorithms. So what I'm saying is you won't gain much from looking at verification stats right now.
-
1
-
-
4 hours ago, Froze were the Days said:
The turd that is the CFS currently similar at 120z to other outputs but drops the trough to the north over us (again) and fills in situ, so no easterly.
CFS is not designed as a model to accurately model the next few weeks. The point is to let it run out into the far distance in order to spot any strong long run trends. It did pretty well with spotting the colder than normal synoptics for late December. Not to say it has Nostradamus like predictive abilities.
-
1 hour ago, Sparky68 said:
Parts of America still getting a pasting , not jealous at all!!!!!
I spend a large amount of time, due to work, in LaCrosse Winsconsin (next to Minnesota). It gets to -30c to -40c in the heart of winter. It's too cold to be outside and quite dangerous. So there is such a thing as too cold! The Mississippi freezes over and you can go ice fishing in your car.
-
1
-
-
4 hours ago, slater said:
You are correct, just watched the the Met Office 10 day tread, they have said yes it will be a easterly next week but the source is from the med where temps are in the mid to high teens. Anything that will fall out of the sky will be rain. As I said the otherday the north sea is still at 9c anyway.
You want the sea to be warm but the uppers (850s) cold to get good snow of an easterly. The issue is the uppers aren't cold enough on this run.
-
1
-
-
55 minutes ago, Trom said:
Well unless the models has significantly improved since the last major SSW then you are correct. Every time in the past when there's been a SSW event it's thrown the NWP models into a tail spin. It's normally taken a few days post event for things to be resolved (if my memory serves me correctly). Lots of factors in play:
1. Speed of the down welling
2. Location of blocks and cold in the mid lats
3. Most have happened when we've been locked into relatively typical weather. This one happens at a time when the trop vortex is anything but normal
So interesting times but I wouldn't think any model has a handle on this currently (again only going by past experience).
Just to add. The 2018 event was a very quick response coming about 2 weeks after the event. It can be 1-2 months. Having said that it looks like this one will be reasonably rapid.
-
Minus 3.3 in South Croydon
-
1 hour ago, Catacol said:
That’s a 14 day reversal. And 6 days of substantial reversal. I don’t believe NWP will read the impacts of this well. Expect a lot of model volatility in the days ahead.
Well unless the models has significantly improved since the last major SSW then you are correct. Every time in the past when there's been a SSW event it's thrown the NWP models into a tail spin. It's normally taken a few days post event for things to be resolved (if my memory serves me correctly). Lots of factors in play:
1. Speed of the down welling
2. Location of blocks and cold in the mid lats
3. Most have happened when we've been locked into relatively typical weather. This one happens at a time when the trop vortex is anything but normal
So interesting times but I wouldn't think any model has a handle on this currently (again only going by past experience).
-
6
-
-
4 minutes ago, Paul Sherman said:
So a lot of talk of why are the BBC forecasting a Rain/Sleet Streamer for next week - Have taken a look at 5 of the Models that go to T120 for todays update and will put a percentage in here of what it is currently showing for Monday 4th Jan
Arpege currently showing Rain
Icon currently showing Rain
ECMWF currently showing Rain/Showers
GFS currently showing Rain or Wintry Showers
So at the moment the percentage chances of Snow on Monday are at 0% out of 100 so there is really only one way we can go on tomorrows Update. I will add in the Precip Models as we get closer that only go out to a few days ahead and average out the chances to get the percentage but currently not looking great
Mondays Snow Risk Currently at - 0%
The Tuesday looks a little more promising though in terms of both temperature and precipitation.
-
3
-
-
Temps and dew points gone negative in South Croydon. Temp -0.1 dewpoint -0.6
-
2
-
-
1 hour ago, CreweCold said:
Looking to me like the strat displacement may shuffle the cards against us. A split would be preferable I think.
At this moment there are signals for such a warm up in mid Jan. Longer range extended EC and GEFS ensembles going for it. The question is a) will it happen and b) can we keep it temporary if it does.
In the past a SSW has caused the models to break down for a period. They just haven't been able to model it at first and it's caused a lot of volatility in the NWP output. So I don't buy the idea that any of the models knows the outcome right now. It's been 7 years since our last big favourable outcome and whilst programming and models algorithms have been updated, these are rare events and huge difference from the climatical norm. I don't think there's enough data for them to diagnose all of the impacts. So I'd expect a period of uncertainty based on the following factors:
1. Split vs. displacement
2. Speed of propagation from strat to trop
3. Location of the blocks in the mid latitudes
The impact of these events is relatively new to the modelling process. It's an area that had a thread tucked way away from prying eyes even in these forums until recently (last few years).
The understanding has come on leaps and bounds recently. Very unusual to see one when the trop is already disturbed from its norm.
-
1
-
1
-
-
3 hours ago, bumpydogz said:
50% of them maybe?
That would be the median not the mean
-
1
-
-
1 hour ago, Mike Poole said:
I think there are 3 reasons for this. First, the evidence, from the verification stats the GFS is not as good a model as ECM:
At day 5 it is behind the ECM, UKMO and just behind the GEM (CMC on chart).
Secondly, it runs 4 times a day so gets more airtime. Thirdly, all its output is freely available, whereas only limited output is available free from some of the other models. So you see more output from GFS on here for those reasons.
Presumably these are verification stats for the entire northern hemisphere. I always assume (possibly incorrectly) that they would be much lower for the UK in isolation, given we are sat at the edge of an ocean with a continent to the east.
-
1
-
-
2 hours ago, Mike Poole said:
It is a new development from HIRLAM apparently, this explains:
http://en.wikipedia.org//static/favicon/wikipedia.ico HIRLAM - Wikipedia
EN.WIKIPEDIA.ORGI'm pretty sure it's been around for a while (at least last year)
-
1
-
-
3 hours ago, Mike Poole said:
Looks like a direct hit from the north on the JMA T192:
Good 12z output today, confirms the view we are in the game, and probably will be for a few weeks if the strat and trop vortexes stay disconnected, actually getting more so on the NAM plot, based on todays GFS 0z:
Thereafter, the scene seems to be set for a possible SSW with the Aleutian low, Russian high combo, a significant precursor, so recurring in the output, illustrated on the ECM T240 chart:
With the start of winter only 4 days away, coldies could hardly have more irons in the fire, so to speak...
We would be wanting a slow response from any SSW at the current point. Right now I'd like to see how the current scenarios play out. So much more promising than in past years. I always visualise that a SSW is like breaking at pool - bits of vortex are going to fly but exactly where and when is always the issue. Like a reset button. If the trop and strat vortexes couple any time soon I'll be praying for the SSW, but coupling doesn't seem to be on the cards early winter. Traditionally snowy weather before Jan gets progressively less probable the further South you go, largely because the NW feed doesn't have bite. For those of us in the South the Easterly is king and you can't expect a developed cold pool just yet. Personally I'm excited - it's only late November. I think the most important thing is to remember there are no guarantees even when the odds look better.
-
4
-
1
-
-
5 minutes ago, mb018538 said:
Shockingly bad in terms of an ensemble outlier yet again. Such a big proportion of gfs runs recently have been either one or the other. Hardly ever representative of the other ensemble members.
The operational is run in high resolution out to 10 days. The ensemble members are not. If you want to compare the model in the early (<10 day) to the ensembles you'd be better of looking at the control which uses the same initialisation data as the op but runs in low resolution. The impact of high and low resolution on the model can be gauged by comparing operating to control.
-
15 hours ago, SLEETY said:
Everything looks dodgy after 96 hours if you ask me, ecm so different in its latter frames compared to 12 hours ago that's its laughable
you can see how the flight data issues are causing problems with the models.
I notice you've been mentioning the lack of aircraft data in the data assimilation stage is affecting the accuracy of the NWP.
This looks fairly logical, but I didn't know how much of an impact it would have, or the amount of input data that comes from this source. After a bit of hunting on the internet I found a study on the accuracy of the modelling (RMSE) when removing each source of input data (admittedly the focus is short run predictions). Aircraft data is most significant for wind and temperature and the study finds it to be significant. So my conclusion is that it must be a significant source of error in the modelling.
Sorry couldn't get the link to the study to work. The title if you are interested is:
Observation System Experiments with the Hourly Updating Rapid Refresh Model Using GSI Hybrid Ensemble–Variational Data Assimilation
-
3
-
1
-
-
11 minutes ago, North-Easterly Blast said:
It just shows how little of note in terms of cold there is in the model output when people are looking for odd ensemble members at over two weeks away, which are never likely to verify as shown, as if a perturbation is wrong at a week out, then it is even less likely to verify at two weeks out.
Given the ensembles have their initialisation data tweaked (control excluded) you could argue that they are wrong from T+0.
To my mind the ensembles should be used to give more substance to the operating run. If the ensembles and op follow a similar path then you can take the op as being more reliable. In other words tweaking the initial data had little impact on the output. Also often you can pin-point the time where the ensembles and op diverge giving you an idea of the timescale that uncertainty is creeping in. The greater the spread becomes the lower the confidence in the op becomes.
The other big use of the ensembles is to look for clustering to spot potential trends.
The other point is that the ensembles are run in lower resolution than the operating run. This difference explains divergence between the control and op which both are run using the same initialisation data.
Other than for fun picking an isolated ensemble run adds little.
-
1
-
-
My god this is getting frustrating - I'm not even concerned about cold right now, I'd just like a little dry. Every time the models seem to show high pressure for the UK it's out in FI and disappears in the next run. Even more frustrating given a pretty unorganized trop polar vortex. My god if we could get some amplification of the pattern rather than this on onslaught of low pressure. This must be 3 months of a relatively similar pattern. Please god just a little high pressure. Walking the dog is becoming a miserable experience. Any kind of block - I'm way beyond caring about cold right now. I can't see anything convincing in the models to suggest it's going to stop. Even the strat is not giving much hope - possible late Jan warmings but that's it.
-
1
-
-
7 minutes ago, abovefreezing said:
I've always found NW radar to be somewhat 'rampy' over the years.
I've just found this https://www.metcheck.com/WEATHER/precipitype.asp?V=Z
How cool is that? Not rampy, either!
I use that site too. The radar is good.
They also use a combo of GFS and other models (albeit free ones) for their forecasts.
-
1
-
-
10 minutes ago, Buzzit said:
I used to do a lot of work in Maclean Virginia (just outside Washington DC) When it snowed there it was epic ! I have just been out for a walk around Chiswick (west london) and the air felt icy cold so perfect for snow
My head office where I have to visit every few months is in LaCrosse Wisconsin currently -24C just before mid-day. Remarkably that's a warm-up for them. A lot of the time you just can't be outside as it's too cold. Issues with frostbite on exposed skin and cold air affecting the lungs. Mississippi freezes and people drive on it and go ice fishing.
-
1
-
-
Wow big upgrade overnight for snow in the South London/North Surrey area from the Met. Double flake symbols instead of single and extension of the period at both ends. Was 6pm to 4am now showing as 5pm to 8am. I have my finger crossed that they're right.
-
1
-
Model output discussion - Jan 3rd onwards
in Forecast Model Discussion
Posted
I may be getting confused but doesn't that paper just focus on the ability of the model in predicting the SSW and not on what the implications post SSW on the trop are?