Corbyn - Winter Warning Of Exceptional Uk Cold And Snow

[asjmcguire]

I haven't seen the forecast, can you just clarify if there any specific figures given that define 'extreme cold'? Does it look like it is in relation to other years at the specific period given, the temperatures we've had in the last few weeks, or is it compared to Winter low temps as a whole in previous years ?

I'm only trying to establish if there is a lower, specific figure where the forecast can be analysed or if it is open to interpretation and judged a success or failure depending on ones own criteria and the general term 'extreme cold' - also against what we've had recently, in past years at the time period given, or just generally Winters gone by?

I'll leave this here for an hour....

Yes, they did do a 'u-turn' a big one, was it last winter?? I remember that they had forecasted a warm winter until the last minute and then revised it.

and then boasted in the news how they had accurately predicted where snow would fall and how much -

despite getting it completely getting wrong

and causing central scotland to shut down for a few days since they claimed much less snow would fall than actually did.......

Edited November 2, 2011 by Paul
Removed protected material.

[Met4Cast]

I'll leave this here for an hour....

and then boasted in the news how they had accurately predicted where snow would fall and how much -

despite getting it completely getting wrong

and causing central scotland to shut down for a few days since they claimed much less snow would fall than actually did.......

Well, that's already wrong going by the models.

[beng]

I haven't seen the forecast, can you just clarify if there any specific figures given that define 'extreme cold'? Does it look like it is in relation to other years at the specific period given, the temperatures we've had in the last few weeks, or is it compared to Winter low temps as a whole in previous years ? I'm only trying to establish if there is a lower, specific figure where the forecast can be analysed or if it is open to interpretation and judged a success or failure depending on ones own criteria and the general term 'extreme cold' - also against what we've had recently, in past years at the time period given, or just generally Winters gone by?

Hi Coast - I've only seen that headline alert for the December period - nothing more than that. Unfortunately you have to pay for Piers' forecasts, so I'll stick with Netweather, GFS and ECM.

[asjmcguire]

Well, that's already wrong going by the models.

But last year the models at this point were not showing any hint of cold either - it was literally an overnight thing when the models suddenly all started showing the cold coming in.

It was explained by those more technical people who understand how the models work that the models are biased to mild winters as that's what statistically we are more likely to experience.

If you remember last year when the cold kicked in - FI was always trying to show the quickest route out of the cold and back to mild. This went on for weeks before we did eventually get back to mild - but not the way the models wanted to do it - it wasn't -15'C one day and 15'C the next.

[Paul]

Not sure about the above - models aren't biased in that way, and there were a lot of indicators for cold prior to this point.

[Aaron]

I'm just wondering what this forum would be like right now if it was 1978.

[Paul]

Probably slower, with less pretty graphics and less members?!

[Coast]

Hi Coast - I've only seen that headline alert for the December period - nothing more than that.

Fair enough then, like many things it will be down to interpretation I guess. I was hoping there might be more guidance to what the extreme low is in the forecast, as even 'normal' end of November temps would be a bit chillier than what we have now.

[johnholmes]

It was explained by those more technical people who understand how the models work that the models are biased to mild winters as that's what statistically we are more likely to experience.

none of the models has any bias built in. All synoptic models simply take the actual data and run from that.

[recklessabandon]

Not sure about the above - models aren't biased in that way, and there were a lot of indicators for cold prior to this point.

I recall upstream indicators for cold being there from early November for sure....but in response to below....

But last year the models at this point were not showing any hint of cold either - it was literally an overnight thing when the models suddenly all started showing the cold coming in.

It was explained by those more technical people who understand how the models work that the models are biased to mild winters as that's what statistically we are more likely to experience.

If you remember last year when the cold kicked in - FI was always trying to show the quickest route out of the cold and back to mild. This went on for weeks before we did eventually get back to mild - but not the way the models wanted to do it - it wasn't -15'C one day and 15'C the next.

I also recall quite a long discussion on the 'bias' of the models, though I think it was possibly more related to the way they handled the data, in that there is for more information/data 'experience' (if you know what i mean) for zonal type conditions, and that, the blocking and synoptics that we saw last winter were far less usual and more difficult to handle for the models - therefore out in FI they wanted to revert to zonal type until there was sufficient upstream information to maintain the 'blocked' synoptics in the near-time forecasts.

I am sure there are others out there who will recall this and can explain it far better than me.

[asjmcguire]

Thank you recklessabandon,

that triggers a few memories.

Apologies Paul re Picture -

I didn't think posting a picture with the rest of the forecast blanked off would be any different from saying - he says it will get colder from the 11th.

[johnholmes]

I recall upstream indicators for cold being there from early November for sure....but in response to below....

I also recall quite a long discussion on the 'bias' of the models, though I think it was possibly more related to the way they handled the data, in that there is for more information/data 'experience' (if you know what i mean) for zonal type conditions, and that, the blocking and synoptics that we saw last winter were far less usual and more difficult to handle for the models - therefore out in FI they wanted to revert to zonal type until there was sufficient upstream information to maintain the 'blocked' synoptics in the near-time forecasts.

I am sure there are others out there who will recall this and can explain it far better than me.

sorry there do not have any bias but simply work on the physics of the atmosphere solving the enormously complex equations with equally complex maths but no bias.

If you do not believe me please e mail UK Met and ask for their view.

[Polar Maritime]

I can remember the discussion on model bias last year to, im shure the conclusion was that yes the Models did'nt know what to make of the pattern change and wobbled a bit which is normal.

[Roger J Smith]

I'm just wondering what this forum would be like right now if it was 1978.

struggling to pay enormous phone bills

[asjmcguire]

none of the models has any bias built in. All synoptic models simply take the actual data and run from that.

While they might not affect the UK weather forecasts - NCEP don't agree with you.

http://www.hpc.ncep.noaa.gov/mdlbias/biastext.shtml

[ukpaul]

I'm just wondering what this forum would be like right now if it was 1978.

A bit like this....

Edited November 2, 2011 by ukpaul

[asjmcguire]

A bit like this....

Where's the huge 9k modem?

[BLAST FROM THE PAST]

Cannae argue with that, John...

No they haven't, Fred...But they do do no more to help the cause than all the others. Except that the Met doesn't feel the need to create arbitrary 'windows' with which they can claim better levels of exactitude than are really there...

Maybe Pete its my memory of 'better' forecasters back in 80s? I'll retract the statement then...as its you

BFTP

[Snowmadsam]

Anyway...back on Topic... (sorry folks)

Personally I don't think the models are biased, I like seeing what they offer, they are there for guidance..and if something is consistent then we can use it to help give us a firm picture of the weather- that's what I think the models are anyway...

[skifreak]

Probably slower, with less pretty graphics and less members?!

:rofl:

As for 78 and 79, see this thread I started about the Blizzard of 78.

none of the models has any bias built in. All synoptic models simply take the actual data and run from that.

Not having bias purposely built in is not the same as them not having a bias. The models are a representation of the atmosphere, surely thus the assumptions and compromises required in the numerical modelling mean the models can not be a perfect representation of the atmosphere, these imperfections having the effect over the model run of effecting certain biases?

[recklessabandon]

sorry there do not have any bias but simply work on the physics of the atmosphere solving the enormously complex equations with equally complex maths but no bias.

If you do not believe me please e mail UK Met and ask for their view.

Given the huge amount of data input variables the models have to process, can they not be subject to bias through omission of a particular variable, or are there safeguards in place to stop that? For instance if an input variable is missed because it is not fully understood or through poor data collation, then over-compensation can occur.

So lets say one area of the ocean has far more data collation points than another area because of the prevailing weather conditions, could there be a risk of overcompensation - because on occasion our weather 'flow' comes from a northerly or north-easterly where there are fewer data collation points leading to over-compensation and potentially therefore variable omission bias?

[asjmcguire]

There is a real world example of this -

You will often hear people say "the GFS is over developing the low pressure"

Frequently this means shortly before entering FI or once into FI the GFS starts reporting

intense low pressure and very high winds.

When / if this does arrive -

the pressure will be at least 10mb higher and the wind will be around 25 - 30mph.

Edited November 2, 2011 by Andrew Simon Jones

[Paul]

That's different to their being a bias toward milder winter's though - pressure differences and the like are typical model biases, they all have them. But there's no default setting or bias within a model for summers to be x, winters to be x and so on.

[johnholmes]

Given the huge amount of data input variables the models have to process, can they not be subject to bias through omission of a particular variable, or are there safeguards in place to stop that? For instance if an input variable is missed because it is not fully understood or through poor data collation, then over-compensation can occur.

So lets say one area of the ocean has far more data collation points than another area because of the prevailing weather conditions, could there be a risk of overcompensation - because on occasion our weather 'flow' comes from a northerly or north-easterly where there are fewer data collation points leading to over-compensation and potentially therefore variable omission bias?

hi

I think you will get a much more comprehensive answer to your question if you google UK Met Office web site, and go to the section on computing. Masses and masses of information in there. I hope that helps.

There is no bias though for any weather pattern, large or small amounts of data, but its all in the detailed explanation on their web site.

:rofl:

As for 78 and 79, see this thread I started about the Blizzard of 78. http://forum.netweat...lizzards-of-78/

Not having bias purposely built in is not the same as them not having a bias. The models are a representation of the atmosphere, surely thus the assumptions and compromises required in the numerical modelling mean the models can not be a perfect representation of the atmosphere, these imperfections having the effect over the model run of effecting certain biases?

not really the atmosphere is fully represented at all levels and all areas-again as with my other reply-try the Met O web site.

I really think some folk are getting too hung up on bias it, it really is not there.

While they might not affect the UK weather forecasts - NCEP don't agree with you.

http://www.hpc.ncep..../biastext.shtml

It is still not bias as the question is being asked-sorry

[recklessabandon]

hi

I think you will get a much more comprehensive answer to your question if you google UK Met Office web site, and go to the section on computing. Masses and masses of information in there. I hope that helps.

There is no bias though for any weather pattern, large or small amounts of data, but its all in the detailed explanation on their web site.

thanks John i'll have a look. in terms of the question I think we (me for sure) were looking at the effect of missing data causing over-compensation in certain areas, especially when the 'ground' data sources we were relying on were from areas where our weather generally does not prevail 'from' therefore increasing the risk of data 'gaps' (IMO) and 'skewing' - probably a better description than bias, which suggests an inclination to one thing or another (almost by design).