Stratos Ferric

Nudging under 3C here, and a bit of sleet in the precipitation at the back of the squall line that just passed through. About 10% snow, about 50% melted. Rain below 200m. Evidence on the tops of a dusting of snow this morning, above around 400m.

Really noticeable drop in temperatures at the cold front night, fairly typically 3C in a very short passage of time, particularly across the south, and the evidence from around Bristol suggests intense downdraughts given that a short spell of snow was reported from quite a few locales in the area.

That would be an astonishing turnaround. The thirty year rolling for Jan is 4.4C: I can't see us ending up anywhere near that mark.

Great post Steve. I recall during the blizzards (and I'm using that word very deliberately and carefully) of 1979 the prisms piling up against the bottom of the window: often the accumulation on the sill would get up to 6" and more. Re conversion ratios the rule of thumb is 10:1, though drier snow may stack more than this (it traps more air and compacts less immediately at low masses) and wet snow rather less. They type of flake also matters: prisms will arrange themselves more tightly than plates, particularly where the snow is wet and chunky. Such is the stuff of avalanche modelling by the way: instability between the layer boundaries brought about by the different crystal structure - all of which further compounds with temperature variation. This is why ski resorts like heavy early snow pre season followed by substantial rain (without washing the base away) - the soaking and freezing helps lock the pack. Like you I recall stading in the snow looking at small plates on my sleeve: until I was 8 or 9 I thought snowflakes were the big gloopy aggregations that we associate and imagine - the reality is more awesome when the air is cold enough to prevent any cohesion. Marvellous.

Jest though you may, I suspect that that outcome may not be far from the truth.

Nowadays it's far better to think about nodrifts. They have been in abundance over recent years. Alas, they come in every colour apart from white. I'm liking your footer Andrew! We may need to draw a map with Abingdon isolines on it soon. I'm fancying that there's something of an epicentre of potential disappointment developing over Ireland.

Yes, I think for me, from here, 3.0C is around the evens mark, perhaps a bit lower. The very interesting baselines are 2.3C - the coldest month in a 'modern' winter, and 2.5C - the coldest January in that same timeframe. I usually take a cut of one GFS run per day (not the same timed run), and the chart below shows the pattern of daily projections based on each run. Certainly consistent still with a landing most probably in the range 2.5-3.5C. I thin you're probably right at the bottom end WiB, but there's little or no chance of us getting to 5.5. We'd need to average around 10C every day in from here to get to those dizzy heights. If the set up was vaguely suggestive of a sustained azores HP there might be a sliver of a chance, but that simply isn't on the cards at present. I think even reaching 120 cumulative degrees from here would be a big ask, equivalent to about 4C.

Yes, I'm feeling slightly happier tonight.

A finishing value of anything above 2.8C would be a big leap by any standards (though less notable in modern times). 5.2C is the largest upwards correction in the series plotted here: I don't see that being breached this year. 3.5C movement would be a top five leap; 3.8C would equal the largest in the even larger teapot (i.e. outturn of 3.6C or higher).

I never knew that John. Why wouldn't they round UP though. I'm assuming the concern is re altimeters, in which case isn't it safer to have the surface pressure reported too high (i.e. equivalent to a lower surface height above m.s.l) than too low. Or is it to support initial calibration (in which case I understand)?

You must have some brass neck given the statement you make in your sig line. The UKMO will have some rational and robust basis for making their projection. I look forward to hearing yours.

TM, you might want to clarify: isn't that a non-sequitur? Either seven days don't exceed 2.9C, or one of the seven does. Or should the first 'max' be 'daily mean'? Maybe I'm missing something... Good stats by the way.

Define "better". One might argue that we should measure as accurately as possible, but if the basic reading is only at 1d.p., than any derived calculations can only, legitimately, be stated at that degree of accuracy. If the base data (daily readings) were made at intervals of 0.01C then in any final report 0.01C would be the best we could report at. In athletics, where these things matter, the timing at the tape is available in 1/1000ths, but the final result is reported in 1/100ths. This is the very best treatment of all for there can be absolutely no risk whatsoever of rounding errors in the final result. If you want me to clarify I'm always more than willing, all you need to do is ask.

Correct, although even the error is subject to a normal curve. Hence, the fact that yesterday the value might have been 4.17C (and therefore a round up to 1 d.p.) has absolutely no bearing whatsoever on whether tomorrow will also be a round up. It's exactly the same as a standard coin toss test.

I think I know what you mean, however you confuse two things. On the one hand there is an error of false precision. On the other, there are very small variations, but variations based on literally millions of data points within which the statistical significance is huge, and beyond debate (i.e. the climate is warming - one might choose to argue over cause). Statistical significance does not rely on the size of a difference, just how big a difference is relative to the normal, or periodic, variation in the base data. I have clients whose machinery runs output volumes in the millions per day, with astonishingly little variation, such that even a small change is almost always indicative of mechanical breakdown.

It's only ever going to be +/- 50% of the next level of unit. So, if we are using values of 1d.p. to derive outputs at 2d.p., the error would be +/- 0.05. Hence, the 9.96 debated here is actually 10.0 +/- 0.05, making the possible range of values 10.05-9.95.

Perhaps, but sometimes these things are amusing. We can all be guilty of taking ourselves a little too seriously, particularly if we get TOO engrossed in the model thread. Skimming the last few days is a right laugh. King Canute had nothing on one or two on here... I can't be bothered to reconcile all that the have done, but see my response above re levelled rounding: this prevents a bias being introudced by always rounding the mid point one way or the other. What can seem like inconsistency and innacuracy MIGHT be very much the reverse. We need more like you when it comes to measuring and reporting no depths. I don't like these new fangled Kentimetres, and with more like you AFF we could jolly well see them off.

Not really, however, this year it matters a bit more than usual. Let's get one thing straight from the off. As far as I'm concerned the 2d.p. treatment is, strictly speaking, false precision. That said, were they to round the two d.p. values to one d.p. (the correct treatment) THEN what the final value hinges on is HOW the rounding occurs. Three of the values COULD be round up: March, May and July. March is indisputable by any accepted convention. May and July cause a problem. On the basis that I was taught at school the 0.05 value would round up, however, as has been debated on here before, a better approach - as used in stochastic modelling - is to round up one some occasions and down on others: on this basis only ONE of the values for May or July would go up. That extra 0.1 makes enough difference for the final outturn of the year to be 9.96 or 9.95, and thereafter one MIGHT argue that the year is legitimately either 10.0 or 9.9. The latter would be a source of unalloyed glee for one or two inhabitants of central London.

As I think I said yesterday, I suspect even that is well out of reach. Sub 3 certainly now needs a major turnaround. I anticipate we'll be above 2.00C by the beginning of next week.

Nick, sorry, I'm being a bit mischievious but also very deliberate. We've probably just had a year below 10.00C, but not below 10.0C, for all that AFF seems to believe we might have. Maybe the weather Gods will deliver a much cooler year this year and then one or two on here can take delight in reprising yet again my previous assertions. They are of course the biggest, worst, assertions ever made, and the only incorrect points of view ever posted on N-W, and it was clearly crass of me ever to have been so strident in what, at the end of the day, was just my point of view. Quite why some people take such gross exception to others' views is beyond me, given that those views are doing no harm. If you want daft go see the model thread.

Great to see that you've not lost your sense of humour Stu. When I last checked we still hadn't had a month sub 3.0C, nor have we had a year below 10.0C. Even if we do have, what of it? I gave an opinion, I might be incorrect, so what?

Strictly speaking that's an incorrect treatment. Averaging to 2dp's numbers stated at 1dp overlooks the margin of error present in the latter. At 1dp values are inherently +/- 0.049 (e.g. a reading of 1.3C might actually be anywhere between 1.25C and 1.34C), thus any average will be +/- the same amount. Hence, the gap between the values for the months you cite is well within the margin of error; either one might, in fact, have been the colder. We had the same discussion two or three years ago when one or two were looking at fractional decreases in annual CET as indicative of an end to warming (that before 2007); the safe rule when averaging is always to move to 1sf less than the raw data, unless that data has itself already been rounded from more accurate reading.

Quite agree re the sustained cold, however, and taking a nudge from earlier references to 1978-9, that year Jan 1 - Feb 21 averaged 0.04C. January that year included six days where the CET was -3C or lower, and only 2 days in the entire month at 4.0C or more. I have always said we will still get some long cold periods, but when they occur they are not as long, or as cold, as those that went before. Finally, there may always be freak events, though I tend to think that excepting external forcing (and events in the early 1980s need to be seen partly in this light given Mt St Helen's) these are highly unlikely re sustained cold (c.f. say, the 1987 storm discussed elsewhere at present, or freak rainfall events like summer 07) It seems quite possible that 1996-7 is still going to be the recent cold winter benchmark after this year. I agree with your challenge, but I take Mr D's point as being that 50 days averaging sub 3.0C in the modern climate is notable: hard to disagree statistically. Your point I think aligns with my general hypothesis that cold isn't what is used to be. If you have no long frame of reference then this recent spell feels very cold, and whilst there's no arguing that it has been cold, it's nowhere near what we used to have even fairly recently. Those who dislike the term "even larger teapot" nota bene.

I didn't draw that by hand you know!

When I last checked the month used by Hadley and Manley is the calendar month. One of the reasons why 3C (or any extreme you care to mention) is hard to land is not because it applies to ANY thirty day period, but to a specific period. What has to be remembered if you start to cherry pick periods is that it's swings and roundabouts. Say you had 1-15 of Jan average 10C; 16 Jan -14 Feb average 0C and 15-28 Feb average 10C again, the record book would show a pair of 5C calendar months. Some on here would be pointing out the 0C period, but in so doing would overlook the fact that the mean of the remained is now 10C. All other things being equal, you're about thirty times more likely to land a cold thirty day period than you are a cold calendar month, howsoever 'cold' be defined. My argument has always been that landing a 3C month is probably the modern floor, and that each year that this doesn't happen makes the future occurrence less likely. In practice, and at risk of appearing to be side stepping, the floor may be slightly below this (3C is a convenient round number), but the point in a warming world - and hence my reference to Jan 1997 - is that every time we get a cold month the value reached is a tad above the previous such low point. At the other extreme new records for warmth are being set, AND against recognised calendar periods - not just thirty day periods of convenience. January 1979 also had snow on the ground for most of the month over most of the country. We are NOT in 1979, any more than we're in 1947 or 1963. Presumably you'll also hold the same view (re people surpassing themselves) re many of the winter LRFs then will you OP? Quite why people have to react to any comment I make I really don't know. I state my point of view; agree or disagree, but let's not make out that to have a point of view is a bad thing. Saying 1 - 5.5 C is possible is a bit like saying "I don't know what day tomorrow is but it's either Sunday, Monday, Tuesday, Wednesday, Thursday or Friday". If we come in below 2.0C I will personally eat EITS.