Roger Smith's winter forecast 2010-11

[Roger Smith]

This is the unvarnished copy from my late September post on a forum that shall not be named ...

Winter 2010-11 -- Mild to start, cold to finish (for the eastern U.S.)

____________________________________________________________________________

This forecast has been based on my research model and the index values generated by analogue components from 170 years of temperature and precipitation data for Toronto, Canada and extended to the scale of the continent by some other index values for monthly temperature anomalies and application of general theory being developed.

I could mention my summer forecast published here separately as some indication that my research is capable of producing an accurate seasonal forecast, but as to winter forecasts, I have had some ups and downs in my first two attempts here, and looking back over a longer period, I find that the general winter outlook is running a little better than 60% accurate using these methods, so caveat emptor. I was going to wait until later October, but as there are so many forecasts out already and I find my index values similar in trend to quite a few, I thought the main value of the forecast might be on the daily scale where this method does give results (people may remember the triple phaser forecast for Jan 30 2010 that actually turned out to be the first of the six or however many major snowfalls on the east coast).

I should mention in passing that I do not use a "la Nina climatology" approach because I find that (a) there is too large a range of possibilities, but also ( such teleconnections are presumably built into the alternative approach being used, and I can say that my research model for predicting the SOI gave a hit on the 2009-10 El Nino and now on the 2010-11 La Nina, and goes on to show a very strong El Nino in 2012-13. Quiet sun has been factored into the forecast subjectively by dropping the anomaly trend curve by 1 C deg. This makes very little difference to the pattern forecast and just tilts the balance more towards cold at the end of the winter.

Overview

The hemispheric winter may be dominated by a large and intense cold pool centered on eastern Siberia and extending across the Bering region into northwest Canada. Some retrograde episodes in early winter may try to build a Greenland blocking high but this will not be as strong or persistent as in 2009-10.

For eastern North America, the late autumn should turn mild again, once the current cold trough regime fills, and positive anomalies are expected from about Oct 20 to Dec 15 in a zonal type circulation pattern with strong inflow of Pacific warmth and moisture (the latter mainly exhausted by orographic snowfalls in the west). This pattern should begin to shift late December to a modified colder broad trough over central to eastern North America with a range of daily conditions averaging near normal in the east, while the western interior becomes much colder than normal with the advance of the extreme cold from the Yukon trans-Siberian source. Then later still in the winter, a deeper trough is likely to form near the east coast with a good chance for a modified repeat of last winter's heavy snowfalls but this time probably extending further north due to the absence of the mild northeast flow component. During February the western half of the continent is likely to warm gradually through the growing influence of ridging over the west coast and Rockies, forcing the arctic highs in northwest Canada to take a route through Manitoba and the western Great Lakes.

Overall, therefore, the DJF anomalies are likely to be slightly negative in most regions, trending to near normal in the southeast U.S. and highly negative in central Canada. But the monthly anomalies are likely to be something like +1, -1 and -3 F for the three months in the eastern U.S., while around +3, 0 and -4 in the central U.S. and -1, 0 and +2 on the west coast.

Precip will be generally near normal but snowfall will start out being rather subnormal in many regions and grow to become above normal in February. Would predict seasonal totals of about 25" for DC, 32" for PHL, 40" for NYC, 65" for BOS, trending to 80-100 inches inland parts of PA and NY, New England. This should be an average winter for lake effect snowfall and the WSW wind component that favours Buffalo and certain other zones like Bruce-Grey counties of Ontario, Kingston ON and north of Watertown NY etc, may be more in evidence than west to northwest types.

Some daily details

Within this pattern (which seems to be at least partially consistent with other LRFs I have seen) based on the energy-level model within my research, I can offer some possible daily-scale events that might give some flavour to this seasonal stew. These would be highlights of the evolving season rather than any effort to give a daily-scale forecast for the whole winter. As an overview, there are three sets of energetic events identified in the model, and timing lines (where these events should be located at event times) that run approximately (a) through the western interior U.S., ( from the western Great Lakes to the Carolinas and © off the east coast of Newfoundland. So at each of the major energy peaks, substantial lows should be located somewhere near these three timing lines, then they would move east to transform into the next series of events in the model series. There are some weaker energy peaks taking some of the available time spaces, so that this following scenario is just a series of highlights rather than the whole forecast scenario. I will concentrate on the events for eastern North America and just mention a few of the western or far-eastern events in passing.

In December, the new moon event (Dec 5) should be a mild, zonal type low running across the central Great Lakes. Around this time, the northeast U.S. should be in a showery, mid-40s southwest flow, trending to dry and 70s in the southeast. This is followed by another strong wave around the 7th that would probably introduce a modified cold spell for a few days.

The next major energy peak occurs Dec 12-14 and should be more or less a reload of the above. Both of these energy peaks will likely be times of heavy rains and strong winds on the west coast with heavy mountain snows developing.

The full moon and N Max energy peak on Dec 21 could produce a major storm in the Great Lakes region, tracking from about KS to IL to s/c ON. This may be the first major snowstorm of the winter for the western Great Lakes (timed for Dec 20). It should remain quite mild in the eastern U.S. and could become near-record warm in places, with moderate to heavy rains spreading up the east coast. This mild spell would likely last into Christmas Day with some tendency to turn colder afterwards. On the west coast, this event may be a major flooding rainfall event from San Fran north to Vancouver Island. Timing would be in two waves Dec 19-20 to 22-23. Heavy mountain snows can be expected and severe cold will be pressing south between the west coast storms and the Great Lakes low.

A somewhat colder pattern in the east may give an opportunity for a 3-6 inch snowfall event around the last few days of December, and lake effect snowfall may be kicking in during this period.

The southern max and new moon events of Jan 2-4 will bring a good chance for another Great Lakes snowstorm and the track on this may run further south, like KY to PA, bringing snowfalls into some of the interior northeast. This will probably stay too mild for snow in the northeast coastal regions, although there may be an offshore secondary making this a rather chilly rain or sleet type event for some. A colder week will follow with some of the more intense cold from central Canada making its way south and east.

The first coastal storm of the season may be at the Jan 9-11 energy peak, although I suspect February will be the best month for these again. Somewhat colder air will continue to mix into the fast WSW flow mid-month and it may be quite squally in the Great Lakes for a while, then a massive arctic high should develop before the Jan 17 northern max (around 14-15th). Around this time it could be snowing on the west coast and well south in the Rockies also. Intense cold will be reaching its strongest extent over western Canada.

Then we have the auspicious energy peaks of N Max and full moon from Jan 17 to 20, which often signal major storm development. This may be a pattern that favours a high energy storm developing in the Gulf of Mexico moving northeast across the inland southeast and Delmarva towards New England, followed by a coastal storm. I'll just say watch for this period and expect some major snowfall amounts especially in parts of New England and inland NY-PA.

From this stormy period to the end of the month, it's likely to be a rather cold and windy week to ten days with disturbances moving southeast bringing lake effect and widespread lighter snowfalls. The west will continue to have some near-coastal snow and chilly rains as the storm track tries to break down the arctic high inland.

February may be one of those more traditional cold months, not confined to heavy snow in one region as last winter, but with more widespread east coast snowfalls developing. The energy peaks are around the 3rd, 10th, 14th, 18th-19th and 24th of the month. I am hopeful that many will see some 5-10 inch snowfalls if not better, during those five storm events, as the storm track is likely to remain something like TN-s VA-Cape Cod. If you get lucky, this could be a 2010 repeat for some and extension for others.

These would be periods for heavy rains on the west coast as the storm track lifts and starts to hit B.C. rather than OR/WA. The chinook may develop at times over the Rockies and the cold core will shift from Alberta to Minnesota during the month.

March is likely to continue colder than normal trending to near normal second half. There could be several more snowfalls during the cold period, and then the "Ginx gale II" on March 19th at the perigeean full moon. I'm speculating this may be a deep nor'easter with some mild air wrapping around near BOS and PVD but possibly a heavy snowstorm from there west and north.

Let the fun begin !!! (snowmen x 3)

-----------------------------------------

Since I posted that, have worked out a European forecast and warned folks in the UK and Ireland on weather forums that they might be in for a stormy, at times anomalously cold and snowy, other times windy and wet, winter season, with a super-charged storm track trying to stay south of its normal position.

Watch for the major storm windows of opportunity about every 7-8 days starting with the next full moon, and peaking around the full and new moons through Dec and Jan, as well as intermediate dates from a second set of energy peaks.

The peak on December 21 looks particularly promising as full moon and northern max will occur simultaneously there. I expect a storm track through the upper Midwest to bring a blizzard to parts of KS, NE, IA, SD, MN and possibly a bit further south but this may be a mild storm for the lower Great Lakes and east coast. It would stand to reason that if this storm cuts north, a very mild spell will occur Dec 20-24 in the east.

[MJO812]

40 inches for NYC

Nice forecast. Hope it verifies.

[weathervswife]

OK RS, KA, and well respected Phily met RG having a snow party. Why are you people giving me hope? I had resigned myself to a year off.

[Ginx snewx]

March is likely to continue colder than normal trending to near normal second half. There could be several more snowfalls during the cold period, and then the "Ginx gale II" on March 19th at the perigeean full moon. I'm speculating this may be a deep nor'easter with some mild air wrapping around near BOS and PVD but possibly a heavy snowstorm from there west and north.Hey thanks for the shout out but it will be Ginxys Gale 3 , good luck with your forecast.

[Roger Smith]

I think you're right on that one, and in general I expect some intense storms to develop this winter, rather than one of those flabby La Nina type winters where nothing much happens, I figure this will be a high-energy winter with deep lows that will keep the forum hopping.

[snowlord81]

Nice forecast Roger, I remember last year you did pretty good calling storms from a long ways out using your method. I know this was more East Coast oriented, but what are your thoughts in terms of the midwest and ohio valley temp wise and snowfall wise? It obviously sounds like your predicting an active winter east of the mississippi. Thanks for your thoughts.

[Roger Smith]

Well, I wish I could do maps as well as the other long range guys, working on it ... but in words, I think December will be a case of a sharp temperature gradient from very cold in central Canada to very mild in the southeast US, so with the storm track something like OK-IN-QC expect the rain-snow boundary to lie frequently just north of that and temps to range from maybe +2 in southern IN to -4 in s MN and -8 in ND. In January that storm track is likely to start sagging further south and while I think the east coast may hold on to near normal overall the month will likely turn colder, and the midwest therefore may have a relatively cold month leading to a widespread cold February east of the Rockies, while the west warms rapidly as the cold is shoved east rather quickly by ridge building over the Rockies. You can see my snowfall estimates in the Beau snow contest thread. I think I've gone for generally above normal snowfalls in the Midwest. I'm starting to think 1971-72 might be one of the better analogues, perhaps a more energetic form of that winter, and also some similarity to 1995-96. If we're going to do a triple phaser this winter, it would most likely come around Jan 20th.

[snowlord81]

Well, I wish I could do maps as well as the other long range guys, working on it ... but in words, I think December will be a case of a sharp temperature gradient from very cold in central Canada to very mild in the southeast US, so with the storm track something like OK-IN-QC expect the rain-snow boundary to lie frequently just north of that and temps to range from maybe +2 in southern IN to -4 in s MN and -8 in ND. In January that storm track is likely to start sagging further south and while I think the east coast may hold on to near normal overall the month will likely turn colder, and the midwest therefore may have a relatively cold month leading to a widespread cold February east of the Rockies, while the west warms rapidly as the cold is shoved east rather quickly by ridge building over the Rockies. You can see my snowfall estimates in the Beau snow contest thread. I think I've gone for generally above normal snowfalls in the Midwest. I'm starting to think 1971-72 might be one of the better analogues, perhaps a more energetic form of that winter, and also some similarity to 1995-96. If we're going to do a triple phaser this winter, it would most likely come around Jan 20th.

Thanks for the thoughts Roger. 95-96 was a great winter here in Indiana. Those triple phasers...man I wanna see just one in my lifetime (hopefully on the good side of the track) A 1978 redux would fulfill my needs if only...

[Ender]

This forecast has been based on my research model and the index values generated by analogue components from 170 years of temperature and precipitation data for Toronto, Canada and extended to the scale of the continent by some other index values for monthly temperature anomalies and application of general theory being developed.

This would, at least at first glance, appear lacking in practicality, if not reason. Are you literally basing your forecast on an index derived for Toronto or are you applying that index that you found for Toronto to recent historical data at other locations as well?

I could mention my summer forecast published here separately as some indication that my research is capable of producing an accurate seasonal forecast, but as to winter forecasts, I have had some ups and downs in my first two attempts here, and looking back over a longer period, I find that the general winter outlook is running a little better than 60% accurate using these methods, so caveat emptor.

FWIW, the NWS used to use something like 60%, perhaps it was 66%, as the litmus test for a model forecast. If the forecast was at or below this percentage it was deemed "worthless/irrelevant", perhaps they have a different threshold for their long range forecast as they can't be doing all that well, at least not for Winter/Spring.

You mention you've posted two forecasts. Do you have more examples of forecasts that you haven't posted along with the validation results? It'd be interesting to see if your index works for some seasons but not others...or some regimes (Enso, PNA, etc) and not others.

[Roger Smith]

Well, in practical terms, if you could make a perfect forecast for any point such as Toronto, and had the right general idea about circulation patterns, then it would be relatively easy to extend that detailed approach for the one rather central point to a fairly wide area, but also to answer your question in more precise terms, I have taken the approach of studying less detailed index values for other points on the grid and I have extended this research since 2005 to include the Atlantic and European sectors. Given that the large organizations have billions and I have to raise my own research funding, my practical range is somewhat limited but I'm getting some encouraging results. The summer forecast is over on the old forum and it seemed to verify quite well on the heat waves. That 60% figure is very approximate, and I'm aware of many different ways of assessing LRF accuracy, correct anomaly sign is pretty low-grade validation technique compared to SD or any other more precise evaluation. Anyway, I would be the first to admit that my approach is somewhat hit or miss in practice, I think it contains the right way forward because the better half of the forecasts are really quite good even on details, but sometimes the results are skewed and show that the index approach has not isolated all the relevant factors. What I suspect is that while I have identified about 120 index values and crunched my data bases to generate the current predictive values for those, there may be more than 200 in reality, and I have probably found the larger ones but if enough of the smaller ones are aligned against the composites that I am able to generate, then the results get thrown off. If those smaller unidentified factors are scattered at random then they tend to cancel out and the main index values do a good job of approximating the forthcoming seasonal reality. To give some idea of how fine-tuned this is, even the larger index components, taken alone, rarely display amplitudes of much more than 10% of the total range of temperature at Toronto (or presumably most other places) and the average is about 0.5 C deg or a little less than one F deg. If one tried to use any one of these alone (as I suspect some researchers have done in the past) then the results are bound to be poor. I've managed to identify several groups or families of index values, some of which relate to lunar-atmospheric interactions, but even taking all these related groups together usually gets into the 15-20 per cent of variability range. Taking them all as a sort of composite seems to bring us past random background noise into something approximating the thresh-hold of what people would generally consider reliable or a breakthrough, something that it's worth stating in context is not widely accepted to have been achieved by any methodology, although there are claims out there. I don't make claims that extensive, I make the claim that I am on the right track to a result that I realize is extremely complex and to be practical would take the application of a lot more resources than any individual could bring to bear on the question. And I am not the one who has delayed or prevented that from happening, but those who have considered my requests for development are the ones to answer for that.

[Ender]

Well, in practical terms, if you could make a perfect forecast for any point such as Toronto, and had the right general idea about circulation patterns, then it would be relatively easy to extend that detailed approach for the one rather central point to a fairly wide area

This simply isn't true. It may in fact be possible to make a perfect forecast for a point, and we (collectively) have an outstanding understanding of "circulation patterns", not to mention fluid dynamics, the laws of thermodynamics and the gas laws. Making a (relatively) "perfect" forecast for a point, or a mesoscale region, is within our grasp now. It's simply a matter of enough computing power and dense enough, highly accurate initial conditions and very, very fine time stepping. Maybe it's only a "perfect forecast" for the next 120min, but it can be relatively perfect for your geographical point. Now, let's say that you have an outstanding understanding of the "circulation patterns" in and around Toronto. Does that mean you can extend your "perfect forecast" outside the initial well sampled domain to those areas just 20 or 30KM outside the foretasted area? Sure you can, but the forecast sure won't be "perfect" for those areas, in fact it'll rapidly deteriorate to nearly useless.

I'm probably reading too much into your use of the word "perfect", though...

...but also to answer your question in more precise terms, I have taken the approach of studying less detailed index values for other points on the grid and I have extended this research since 2005 to include the Atlantic and European sectors. Given that the large organizations have billions and I have to raise my own research funding, my practical range is somewhat limited but I'm getting some encouraging results. The summer forecast is over on the old forum and it seemed to verify quite well on the heat waves. That 60% figure is very approximate, and I'm aware of many different ways of assessing LRF accuracy, correct anomaly sign is pretty low-grade validation technique compared to SD or any other more precise evaluation. Anyway, I would be the first to admit that my approach is somewhat hit or miss in practice, I think it contains the right way forward because the better half of the forecasts are really quite good even on details, but sometimes the results are skewed and show that the index approach has not isolated all the relevant factors. What I suspect is that while I have identified about 120 index values and crunched my data bases to generate the current predictive values for those, there may be more than 200 in reality, and I have probably found the larger ones but if enough of the smaller ones are aligned against the composites that I am able to generate, then the results get thrown off. If those smaller unidentified factors are scattered at random then they tend to cancel out and the main index values do a good job of approximating the forthcoming seasonal reality.

This sounds like a justification of the ensemble process. Well, all except the lack of funding part.

To give some idea of how fine-tuned this is, even the larger index components, taken alone, rarely display amplitudes of much more than 10% of the total range of temperature at Toronto (or presumably most other places) and the average is about 0.5 C deg or a little less than one F deg. If one tried to use any one of these alone (as I suspect some researchers have done in the past) then the results are bound to be poor. I've managed to identify several groups or families of index values, some of which relate to lunar-atmospheric interactions, but even taking all these related groups together usually gets into the 15-20 per cent of variability range. Taking them all as a sort of composite seems to bring us past random background noise into something approximating the thresh-hold of what people would generally consider reliable or a breakthrough, something that it's worth stating in context is not widely accepted to have been achieved by any methodology, although there are claims out there. I don't make claims that extensive, I make the claim that I am on the right track to a result that I realize is extremely complex and to be practical would take the application of a lot more resources than any individual could bring to bear on the question. And I am not the one who has delayed or prevented that from happening, but those who have considered my requests for development are the ones to answer for that.

I can't prove what I'm about to say, but whether it's 120 or 200 or just four indexes, I bet that most of the any success you're having is the result of a single index or two becoming predominant over the others at least for a moment (days or weeks) of time. So, for instance if you're "nailing" the timing of the one or two 1 - 2wk cold snaps during an otherwise typical La Nina winter...then you might simply (easier said than done) finding the time when both -NAO and +PNA might coincide. While that might not exactly be easy...it's easier than figuring out when 200 indexes might find some harmonious balance. If you're getting to the point where you're seriously deriving 100+ indexes, then you probably should look at building a correlated DB of 500MB, SST, and surface anomalies that can then be used as a roll forward forecasting tool. You could even include the moon and sun. If you wanted.