848 replies to this topic

[Roger J Smith]

I disagree with that, the human mind has already mastered things that are probably more complex than whatever the eventual solution to this puzzle may involve, and I think a solution (within a reasonable error range, this is after all the atmosphere we are talking about) will come, but I suspect it will come through a long R & D program. You don't need supercomputers for this, you need a funded, dedicated group with enough resources to build up at least an adequate regional grid. I'm already convinced that if I had that sort of funding and resources available, then what I'm doing could be expanded to a larger and denser grid of points for generation of similar index values, and this could include more parameters. While a "supercomputer" would be almost a necessary part of such an enterprise, I've found that I can number crunch the 240 years of the CET in about an hour to two hours once I set up the right matrix, and my home computer isn't even state of the art. It amazes me how fast a home computer can generate a full set of averages for some index value that might involve 30 or more terms for each of ten thousand data points. Probably the time required is about a half second, a supercomputer could do the same thing in a nanosecond, I suppose.

Anyway, the truth is probably between the most optimistic and pessimistic statements we have just seen here. There has been some advance in LRF recently, but nobody credible thinks we are "there" yet. The only way to get there is to do research in the way that some of us are doing research. I should point out, while it's valuable to get the input of various observers with their forecasts, some of them are little more than subjective estimates based on reading other forecasts. That might not be such a bad way to do this, after all, that's how business as opposed to science normally operates. But it does not illustrate any principle of "where LRF research is at today" for somebody to read six forecasts, think about it for a day or two, then generate their own blend or estimate. Now if people actually have a research basis for their forecasts, then the results directly validate the research.

One other thing, you can't credibly use details that are in forecasts to refute forecasts. It doesn't make sense to me to read that "nobody predicted all this mild weather" when in fact somewhere in every one of these forecasts are mentions of periods of mild weather. So it's more of a quantitative thing and with two weeks of severe cold this won't be remembered as a "very mild" winter so much as a mild winter with a significant cold spell. In Ireland on the other hand where the cold never really dug in, they would be more justified in saying this was a very mild winter, so that might apply to a few western parts of the U.K. as well.

Anyway, final thought as I don't want to beat a dead horse in this thread, where I do agree with TEITS would be that advances won't come suddenly or certainly easily. Advances will almost certainly come incrementally and it could be well into the decade of the 2020s before there's a general feeling that LRFs are significantly improved. Even if one person (let's say I got lucky and my research started to reduce error faster than I've seen in the past five years) begins to demonstrate an accuracy level that begins to converge on something like right side of normal 75-80 per cent of the time, that will not just be obvious to everyone as all forecast methods edge forward erratically or bounce around between good and bad examples -- and there will be nothing to stop the field from being crowded with hopecasters as we've seen (apparently, or some research is totally blown out of the water after that December business). So let's say somebody (and it could be anybody, since we don't know for sure which research paradigm will yield the best results after 10-15 years) does get to that higher level, it could be quite some time before people in the field generally accept that this has happened, and then there would be the question of what to do about it, because if it's a personal research program and the person is 70-80 years old, then you have very practical handover issues looming.

I'm just going to keep doing my research as long as possible and let nature take its course. Unlike some pessimistic appraisals here, I believe there has been progress and I'm "in the ball park" more often than random would suggest. For example, there was a clear forecast of a very cold December in the previous winter, and some people on NW said the winter forecast before that was the best of the bunch (that thread may be in the archives, winter 2009-10). Last summer I came up with a modified warmth concept when many were going for a very warm summer. I can think of some clunkers back a few years ago, and this January did not really meet the required standard. So on balance, I am encouraged but certainly not over-confident. It's like a shadowy version of a breakthrough. I wouldn't be surprised if one or two others felt the same way about their work -- things are looking up and perhaps we're on the right track. But I would modify what TEITS said about this being too big a subject for the human mind, it's more a case of the total global picture being a very large subject with vast complexity, but top people in all scientific fields tend to be able to master a lot of detail, just think of what top biologists or atomic scientists or organic chemists must know in detail, or astronomers, compared with enthusiasts or students. There's no reason to think the same is not true for climate science, although the AGW episode has tended to give many the impression that this is a field for the intellectually challenged, even there, I honestly think some of these top climate scientists are just going through what many sciences have gone through in formative stages, a dead-end paradigm that has to be overturned and replaced by something that predicts nature better. Such as modified AGW with a two-thirds natural and one-third human causative foundation.

(later edit) ... It just occurred to me that if you visualized a discussion of physics in 1880 compared to 1920, you would see just how transformational the right sort of blend of intellect and paradigm change can become, basically almost everything being taught in university physics by the 1920s was unknown to the top workers in that field in 1880, and who's to say after the fact how dimly perceived any of it might have been to them? You could say much the same thing about geomorphology before and after the ice age theory was accepted. We may well be in the age where most look at a weather pattern and see an erratic boulder (blocking high) as something the stratosphere dropped, where in fact it was something that a magnetic field moved. That sort of thing, it's no more complex to think of boulders floating in rafts and dropping to the bottom of an ancient sea, or ice fields moving the boulders, but somebody had to think of the paradigm and then spend a lifetime developing it and defending it against a lot of hostile criticism (none of which makes any sense to us now).

Edited by Roger J Smith, 22 February 2012 - 09:01 .

[johnholmes]

THE EYE IN THE SKY, on 21 February 2012 - 23:31 , said:

The only advance in forecasting that I have noticed since I started following forecasts in the 1980s is the 7-14 day forecast has improved.

Not necessarily about seasonal forecasts but day to day forecasts show quite a marked improvement say in the 3-5 day range. Somewhere there is a graph showing that the 4 day forecast is now as acurate as the 1 day forecast was 30 years ago.
Other graphs also show, see one below, how much more accurate the upper air forecasts are now compared to several decades ago when computers were first used to model the atmosphere.

[shedhead]

To my mind GP has been pretty much bang on this Winter(as he was last), the main problem being some read his forecast of a 'colder' February as another potential Dec 2010, something he never claimed was likely. All in all it's been a very disappointing Winter for cold away from Central and Eastern England, just take a look at the Scottisk Ski resort cams in what is supposed to be the height of the season, I've seen better, more sustained cover in May than there is now.

As for the others forecasts, the one that sticks out for me is Lady of the Storms', who repeatedly said mild was more likely than anything.

As for my own brief forecast issued on Oct 23rd, not to bad imo.

Posted 23 October 2011 - 22:55 In essence drier than average for most of the UK, mildest in the NW, coldest in the SE. Looking farther afield, dry and warm Iberia, dry W Europe, colder from Poland-Italy eastwards, coldest against the average in SE Europe where there will be some significant snowfalls.. .as ever tho, time will tell.

[bluearmy]

snow? norfolk n chance, on 21 February 2012 - 18:48 , said:

Ian Brown's would have been good, especially for Dec, most of Jan and end of Feb

i was going to post on this as ian did make a point that he felt the p/v would split in jan but that we would be left in the wrong position to advect deep cold over us. not right but not wrong either. infact, given what the remainder of feb has brought, probably closer to right than wrong. what a shame that he cant find a way to communicate without sounding so arrogant.

[Roger J Smith]

The question of long-range forecast verification is not independent of practical considerations about the gradual degradation of forecast accuracy in the medium term. Let's say for the sake of argument that the first three days are now reliable and that accuracy is mainly degraded by subjective human interpretation factors (plus the very occasional case of model mis-handling).

Our operational experience is that forecast accuracy then degrades from say 80% accurate at day 3 by almost 10% a day until we are getting down into random territory around the end of the medium-range time scale (days 10-12). Forecasts for 15-30 days especially if forced into a detailed daily validation paradigm are little better than random now.

This means one of two things for long-range forecasting. If your LRF is an extension of the proven randomness of medium-range technology near its far end, then how can it be anything but random? There is no known mathematical principle of de-randomized improvement of odds in such a situation. However, if the LRF paradigm is increasingly vague at a rate faster than forecast validation allows, then seasonal or monthly forecasting can achieve less random results, but that's a case of apples and oranges, if the same verification is applied then the results are bound to degrade towards random.

However, a second paradigm is that a long-range forecasting research program is unrelated to conventional operational meteorology and concerns itself only with long-range, therefore having no data base of medium-range forecasts to validate. Then in theory it could improve on the odds of medium-range forecasting at the interface. There might never be an interface visible if the long-range forecast always began more than medium-range distance away from the start of season.

My goal is to achieve 80% accuracy in terms of right side of normal daily-scale forecasting at whatever time range the research model can be expected to operate. Since a component of this is somewhat random very long term wandering of a magnetic field grid, I would cut that off somewhere like 50-100 years with the caveat that large-scale processes like solar wind changes on a large scale or volcanic dust veils might interfere, but otherwise, the concept is to "black box" a forecast system that will always achieve roughly the same accuracy whether you apply it a year in advance or (in theory) from a blind data-free forecast situation (never available to any active researcher) in real time. Now that "same accuracy" might today be 60-65 per cent (this is how my validation studies have averaged out) which is still non-random but only about half way to 80% (from 50%, the caravan of this situation is that you can have a program that black boxes out at 20%, for which see earlier threads on this forum).

Now some might complain, you just made that up about 60-65 per cent, but I would invite you to read the forecast and estimate it for yourself. December was about 25/31 for right side of normal, January about 11/31 and February looks like being maybe 20/29 with the errors mainly in the period 6th to 12th. This is based on numerical index output that is not on screen there but translated into verbal forecasts. There is always the odd outlier day in a well-established trend period, for example, I think 14 to 16 January which were colder than the long-term trend. Now for the season that gives us 56/91 which is 62% right side of normal. My North American forecast has achieved perhaps 65% with somewhat different monthly distributions.

Now, the question would be, where in operational medium-range meteorology does model performance on this one (admittedly crude) statistic fall below 60% -- in other words, at what point is there a balance between this theoretical approach and the empirical approach of meteorology as driven by NWP? My research on that indicates the balance point is around 10-12 days, in other words, if every day you took an existing forecast of mine from say 120 days in advance, and the GFS output for 12 days, then on average I would come out ahead of that. The goal is to converge the research on model run performance at an earlier time each year, but the "hump" is going to be around days 6-8, when that happens, we have the sort of convergence that might indicate that the model concept should transform into the research concept past the reliable short-range time frame.

Those are very advanced goals and probably can't be achieved in my lifetime, but I hope to get as close as possible like perhaps a demonstrated 70% accuracy and equivalence to about day 8, to get in position to hand off the research. It may never be possible to reach 80% consistently, even if the concept is entirely sound, the interaction of elements and possible lack of robust signals might keep the upper limit in the low to mid 70s.

I'm also quite aware that some other research, either different in foundation or inadvertently different (same principles, different taxonomy) could achieve better results, may already have done so, and may also have better political success (ability to connect research with establishment). In fact that would seem to be a very reasonable assumption given my track record with the latter.

[Eugene]

Shocking runs tonight for cold weather fans, last time i looked on monday night HP was going to give us something colder at night, now the pattern has been flattened and another very mild snap is on the horizon for early to mid next week that wasnt there on monday night, now in MD they are talking of cold weather two weeks away, when you come on every two days like i do you see shocking reversals that others who watch every single run every day dont see, time some took a step back and saw what is happening with the weather atm, very mild snaps springing out of nowhere in the medium term and any cold remaining in the deepest realms of FI, maybe eventually they will come off........in May.

Edited by Eugene, 22 February 2012 - 21:46 .

[Boar Wrinklestorm]

Roger J Smith, on 22 February 2012 - 08:51 , said:

... basically almost everything being taught in university physics by the 1920s was unknown to the top workers in that field in 1880

Now this is where I draw the line. You can't really be that old , Roger?

Edited by Boar Wrinklestorm, 23 February 2012 - 10:38 .

[CreweCold]

Glad no-one picked up on my duff forecast lol

OK time to face the music......this is from the Archive threads.......went for a cold start to winter that never materialised HOWEVER the theme that runs through the forecast is for a northern jet riding over the top of the HP cell which, in my defence, did occur and has been a common theme of the winter.

Winter Forecast 2011-2012- Average Winter, Dry, Mild End
Factors affecting winter forecast (period 1.12.11- 29.2.12)

• Enso state
• Solar activity
• QBO
• Yearly analogues
• Jet stream behaviour
• Long range model output (CFS, ECM LONG RANGE etc)

So here it is, the winter preliminary thoughts as promised; this particular has proven more tricky to put together than last year’s offering and so confidence is slightly reduced due to conflicting signals. During the past 6 months we have seen a reduction in the strength of the la Nina that was particularly prevalent during the latter half of winter 2010/2011, even so we remain in la Nina conditions- however predictions are for a weak la Nina rather than a moderate/strong offering. I consider this to be a good pointer to a colder than average winter across the UK.

However, less positive indicators include the awakening of the sun which has seen heightened activity over the last 12 months and a propensity of the jet stream to migrate further north than has been the case prior to the beginning of the year and post 2006.

Long range model output at this moment in time appears to support the notion of a dry winter period, this can be seen within long range forecasting models such as the ECM. Another theme appears with average temperatures being displayed across models such as the CFS. This suggests high pressure over or very close to UK shores.

Winter to start off colder than it ends- November into December looks to be the coldest period of the winter, much like 2010-2011, a much milder end to the winter seems likely with an early spring season.

December

Starting off cold or perhaps very cold with a propensity for high pressure around Iceland and southern Greenland with east to northeasterly winds being the dominant direction. As the month progresses, it remains cold with some very cold minima but with high levels of sunshine relative to the time of year. A dry month with rainfall around 70% of average or lower. Less in the way of snow as the month progresses with high pressure, originally across Iceland, slipping south to settle on top of the UK.

CET estimate: Below Average

January

The month starts off cold and dry with extensive night time frosts but predominantly dry. Day time temperatures will be on the cold side though will feel pleasant in the extensive sunny intervals where mist and fog clears. Throughout the month, high pressure will settle over or very close to the UK but with a southward trend and thus allowing an Atlantic feed into the north of Scotland and Ireland. This atlantic influence will increase temperatures and will begin to spread further south during the month as the jet stream powers up over the top of the high pressure cell. An average month in terms of temperature but possibly below in terms of precipitation.

CET estimate: Average

February

An above average month in terms of temperatures- synoptics akin to February 1998 though perhaps not as mild. High pressure slips away south with powerful jetstream to the north. Rain and drizzle commonplace across northern and western areas with the driest weather across the south and east. Limited frost potential.

CET estimate: Above average

These are my preliminary thoughts on winter 2011-2012, the final forecast will be released around the middle of November.

Edited by CreweCold, 23 February 2012 - 16:46 .

[Backtrack]

Hey guys, this was my forecast issued in October:

Backtrack's winter forecast 2011

Hi everyone. This forecast has been created from the trends and patterns of numerous model outputs ranging from the GFS,ECM,CFS and a couple of other factors too. Please don't slate my forecast if it does not show what you would like to hear, I have put a lot of hard work into creating the forecast and think you will find it fairly accurate up until January at the least. At this stage confidence falls drastically, with January looking likely to go either way at this stage.

My confidence in February is surprisingly high, not just based on the CFS anomaly charts' trends, but also due to the fact that I believe we are long over due a cool February. And whilst weather doesn't follow a set of rules, I certainly cannot see this month being mild at the very least.

Hope you enjoy reading it, sure I'm an amateur and you will find that RJF's forecast and BFTP's (when he does his) forecasts are likely to be much better, but hey, it's all in good fun!

Also, on a side note, I am a massive coldie fan, and writing the forecast for the first 2 months was painful for me, not only am I confident in their accuracy, but I am unfortunately expecting little cold for Western areas for some time. I've spoiled it enough. Good luck.


October:

21-30th - This period is likely to be mild and windy at times with the best of any brightness in the East, with the West seeing cloudier conditions with rain, heavy at times. Temperatures remaining around average for the time of year, perhaps slightly above average at times in the far South.

November:

1st - 10th - Numerous outputs, trends and patterns point to a North/South split during the first third of the month. It's looking very likely that it's going to be very average in the North, day time temperatures hitting 8-10C, but above average in the South, temperatures hitting 13-16C. Precipitation wise, the first start of the month looks like being very wet and unsettled, perhaps making up for the drought in place in Eastern areas. An Atlantic dominated first third, night time temperatures unlikely to bring any frosts at all away from the coldest of hollows.



10th - 20th - This period will follow the same theme as the first third for a time at least, with temperatures being pleasant enough in the South for T-shirts to be worn. In the North though, it's again very average for the time of year until the mid month point at least. Temperatures again 8-10C, 13-15C in the South. It's the mid month point that may bring some interest to you with the first frosts of the month likely, temperatures still recovering to average by day. This third is likely to be the windiest with frequent gales, especially in the West.

20th - 30th - Another dissapointingly average period with little in the way of Winter to be shown just yet. Temperatures again remaining oh so average, although temperatures may be down a little way at least, 5-9C in the North, 10-13C in the South. Rainfall should be around average for the time of year, as should sunshine amounts. A cold spell is likely around the end of the month (The very end of the month) giving the first low level snow to Northern & Eastern areas, this may also mark the start of the Cairngorm skiing season. Frosts are also likely towards the end of the month, but Western and Southern areas are going to have to wait to see anything remotely in the way of snow.


CET - 7.7C


December:

The first third of December will be dominated by a North Westerly flow, bringing Western Scotland, Northern Ireland, Wales and Western England their first snows of the season. The cold spell will not be anything remarkable with temperatures still managing to climb to 2-5C during the day, any lying snow is likely to be short lived, with hail showers becoming a very dominant feature for Western areas during the period.

The second third is likely to me average with little snow away from the highlands. Rainfall and sunshine amounts are looking around average, rainfall slightly above.
The Christmas period is a great day to feel the full effects of Winter as Santa makes his way down your chimney to make many boys and girls happy. Unfortunately santa may melt in England, as there is likely to be no snow away from the hills. For Northern Scotland, a white Christmas is looking likely with snow actually falling on the day!

The last 5-6 days may see the return of Northerly winds but confidence in this stage is very low. A North Easterly lasting 4-5 days seems the most likely scenario at this stage with snow showers for Eastern areas in particular.

CET - 5.4C



January:

January is likely to be a cold month. Where as December will struggle to get any decent frosts, there will not be a shortage of these in January as high pressure takes control of our weather for a time at least, bringing some painfully low night time minima, and temperatures struggling to make average during the day. Sunshine amounts are likely to be above average, where as rainfall is looking likely to be below average.
Snowfall throughout this period is going to be a premium.

The second third is when things start to get interesting. We start with winds switching to a morth North Westerly element, and as the week goes on, they switch around to North, followed by North East. This is likely to be the snowiest period, with much of the country seeing a good total of snow by the time the spell has finished around the 20th.

The last third is looking increasingly Atlantic dominated, with rainfall above average, and sunshine at a premium. Snow over Northern hills.

CET - 3.5C


February

February will see Winter in full force, with frequent cold spells and snow for many. Frequent Easterly winds will bring some impressive totals to Eastern areas, with the West having to wait a while to see anything wintry other than frosts.

The second period is likely to keep the Easterly theme going but perhaps snow showers for the West as the wind kicks in, sending the showers over the Pennines.

The last half is likely to be Atlantic dominated with frequent gales and above average rainfall.

CET - 2.9C



Overall an average Winter is looking likely. Nothing like the last 2, with the overall CET likely to be bang on, or VERY slightly below average.

Thanks,
Backtrack.

Overall, I think I did well for a first major attempt.

I couldn't of asked for more accuracy for October, November & December.
I messed half of January up, but I feel that I nailed February?

The CET's were not correct (as expected) but overall, I think I did rather well for someone who is still learning daily!

So yeah, very happy.
Thanks