I’ve been given a link in email today to a public forecast page for July by weather prognosticator Piers Corbyn, which you can investigate in full yourself here. I find his web pages and forecasts hard to read, and even harder to accept any more, because in my opinion, he presents them like a carnival barker with overuse of exclamation points, bright colors, over bolded texts, random font changes, and fantastic claims. It tends to set off my BS meter like some tabloid newspapers do. Here’s his USA forecast for July:
[UPDATE: 7/8/12 – The full USA forecast has been made available by Mr. Corbyn and is available here for your inspection: http://wattsupwiththat.files.wordpress.com/2012/07/usa-1207-july-inc-public-summary-news-page-full-fc-key-usa-maps-and-extremes-slat8a-prod-29jun.pdf ]
Some people say however, that despite all that unnecessary gaudiness, he makes accurate predictions. Because he’s made a public forecast and advertised its availability, urging “people to pass the links on”, here’s a chance to find out if he demonstrates the skill that is claimed.
He made this bold claim yesterday:
“Terrible weather is coming the world over this July so WeatherAction has issued free summary long range forecasts for USA and for Europe…”
He sounds like Joe Romm or Bill McKibben talking about “climate disruption”. Of course, it could just be another July in the northern hemisphere. Here’s the rest:
The USA pdf link is issued today on July 4th to go with the Europe link issued the day before. We urge people to pass the links on.
“We also expect very serious near simultaneous solar-activity driven deluges and stormy conditions around the world during our top Red Warning R5 and R4 periods. Any communication of the forecasts must acknowledge WeatherAction”
– Piers Corbyn, astrophysicist WeatherAction long range weather and climate forecasters
WeatherAction Free Summary Forecast for July USA:-
“Could it get worse? Yes!” – Extreme thunderstorms, giant hail and ‘out-of control’ forest fires’
pdf link = http://www.weatheraction.com/docs/WANews12No32.pdf
(or no links twitpic = http://twitpic.com/a3y28b/full )
WeatherAction PUBLIC warning Europe July 2012 “Off-the-scale” Flood & Fire extremes likely (WA12No31)
pdf link = http://www.weatheraction.com/docs/WANews12No31.pdf
(or no links twitpic = http://twitpic.com/a3p7pm/full )
The USA forecast map he provides is a bit hard to read, since it seems he scanned it in from print…note the dot patterns in the graphics. I present it here from his PDF page.
Here’s his forecast page for Europe:
He lists “off scale” weather in NW Europe is one of the claims. I wonder how one should define “off scale” weather.
As Carl Sagan once said:
Extraordinary claims require extraordinary evidence
So now that Mr. Corbyn has put forth some extraordinary claims, we can catalog here the evidence to support those claims, and revisit the results at the end of the month. I urge readers to continue to post both pro and con evidence here as the month progresses. I’ll put a link to this thread in the WUWT sidebar so readers can add information that might be relevant.
Since Corbyn is a fellow climate skeptic, let’s give him a fair but factual evaluation to find out if these claims hold up, of if he’s simply following the path of some prognosticators of the past, such as Jeane Dixon, who made claims so broad that even a small kernel of happenstance occurrences after the fact were used to justify confirmation of the prediction. According to the Wikipedia page on Dixon:
John Allen Paulos, a mathematician at Temple University, coined the term “the Jeane Dixon effect,” which refers to a tendency to promote a few correct predictions while ignoring a larger number of incorrect predictions.
I don’t know that is what is going on here with Corbyn or not, but since he’s put out an open 
forecast, let’s find out. Inquiring minds want to know.
UPDATE: here’s a video of Corbyn explaining his methods:
Ulric Lyons says:
July 10, 2012 at 3:22 pm
“Of the 59 flood events on my link above, 7 are not close to an R period, 8 are within 1 day, and the remainder are all within R periods. There are a number of R4 and R5 events with no flood events on that list, notably in the warmer parts of March and May.”
That is meaningless if you don’t know how many days in those statistics are forecast to be ‘R’. Looking at June 2012, 60% of days are forecast to be ‘R’. If you include ‘R within 1 day’ it rises to a whopping 90%!
This is why it’s so easy to get misled.
Willis Eschenbach says:
July 10, 2012 at 9:58 pm
“Science only works because nobody’s claims get the easy treatment. Everyone is treated to the same harsh inquisition, including my own work, and that’s how it has to be for science to succeed.”
But the choice of targets is yours.
Regards.
Thanks for your thoughts, Agnostic. I have defined nothing. Piers is the one who is doing the defining. He said “tornado swarms” in the area “south of the Great Lakes” on the first to the fourth. There were two US tornados during that time, neither one in the area where he forecast it. I’m sorry, but that forecast is wrong no matter how it is defined.
Willis, if you were standing next to your hail damaged car, would you smugly point out that there were no tornado’s and thus Piers’ prediction was wrong?
You ARE defining what constitutes right or wrong because you are using an ancillary weather effect and inaccuracy in timing to discount the prediction. Imagine a bell curve, where the prediction that the event will occur lies over one and a half weeks with the period they think is most likely to occur is during the period 1-4. That means the event is most likely going to occur at that point but could still occur earlier or later. It’s massive missing of the point to examine a precise period for a prediction that was made more than 2 weeks earlier and say there was nothing unusual. It’s like saying ‘I’ve looked behind the sofa but I couldn’t find the fire in the room.’
Also, do not keep insisting that the prediction was MADE on the 28th of June. That was when the forecast was updated and published, but initially the prediction was made well before that. From the point of view of determining whether there is anything significant about the SLAT methodology ignoring this not useful.
If a prediction comes out and says a specific type of event (severe storm in this case) is going to occur in early July probably between 1-4, but in actuality between 5-7 I still want to know how it was that such a specific type event could be predicted so far in advance.
How did they know that type of event was going to happen, Willis?
Was it just chance? Does he just make a whole bunch of random predictions and hope that a few of them will be close enough so that he could say ‘see we got that right’ as a commenter suggested recently. That’s possible.
That’s why YOU need to say what YOU consider to be reasonable accuracy especially in the context of long range forecasting, or at least agree something with Piers. Otherwise your just saying ‘He said there would be severe storms with tornados, but there weren’t any tornados so he was wrong.’ and ‘He said there would be severe storms between 1-4 July but they didn’t happen until 5-7 so he was wrong.’
And finally, he does get it wrong at least 15% of the time so this example may fall within that. But is the success rate really 85%? What are you going to consider a success? In my opinion, the more extreme the event or the more specific the characteristics, and the more confident the prediction (in percentage terms) the more you can determine ‘right or wrong’. Maybe a marking system whereby you judge the amount of time in advance the prediction was made, the accuracy of the characterisation, the accuracy of the timing and location. Then some sort of comparison with standard meteorology ought to be made. You could take ‘marks off’ if the initial conditions suggest that type of event was likely anyway, unless the location and timing were very specific. For example, a hurricane in hurricane season, or a snow storm in winter during a La Nina.
I absolutely agree that you should NOT give Piers or SLAT an easy ride. Why should you? But you do need to analyse in the context of the type material you are looking at in order to see whether there is anything useful with the SLAT approach. Right or wrong is NOT going to cut it. From the point of view of long range forecasting, the most useful prediction is severe weather or characteristic weather. For example, The MetOffice predicted a very mild winter 2010 and Piers predicted the coldest winter in 100 years. We know how that turned out.
@Willis Eschenbach says:
July 10, 2012 at 10:29 pm
“Maybe you see “distinct clustering” in that, but I see nothing of the sort.”
We are discussing daily clustering of flood events, one cannot see any daily detail whatsoever on monthly totals.
p.s. The ability to see patterns that do exist, is a valuable survival aid.
@Martin Gordon says:
July 11, 2012 at 12:23 am
“That is meaningless if you don’t know how many days in those statistics are forecast to be ‘R’. Looking at June 2012, 60% of days are forecast to be ‘R’. If you include ‘R within 1 day’ it rises to a whopping 90%!”
R days this year to date are close to 50%, June and July had a lot more than average. 75% of flood start dates on that list are on R days. I’ll go back through it and see how many flood start dates are 1 day after an R period rather than 1 before, considering that river floods take time to build, and I’ll see how many are on the higher rated R periods.
Willis.. “Maybe you see “distinct clustering” in that, but I see nothing of the sort.”
So where are the R4 and R5 periods on that graph Willis?
Willis..
“No, that’s not what happened. He forecast “tornado swarms” for the first to the fourth.”
No swarms but there was one tornado reported, just not confirmed in Ohio at 6:20 EDT.
Piers warned people to expect extreme weather with thunder and hail. That’s what they got.
He forecast tornados (extreme destruction) they got a Derecho (extreme destruction).
Do you think that it matters to the people who lost loved ones whether it was a tornado or a Derecho that killed them? It was extreme weather and Piers forecast warned of
this.
Willis
“He also did not forecast “large hail”, he forecast “giant hail”, and I found no reports of that either.”
I’m not letting you get away with that one. Large or Giant? By who’s definition? Totally subjective unless someone carries a tape measure, then measures the hail, then tries to
find a definitive scale to plot the sizes against, then reports it.
What’s your definition of giant hail Willis?
His forecast was not 100% incorrect as you stated. It only had a couple of errors.
Piers forecast; “Just south of Great Lakes”.
Storm track? Just south of the Great Lakes.
Piers forecast; “Extreme thunder”.
Storm? Severe Thunderstorms.
Piers forecast; “Large/Giant hail”. Word choice dependant on news bulletin or forecast. Large or giant he means BIG hail.
Storm? Large hail.
Piers forecast; “Thundery deluges and local floods including New York state and probably NY and Washington DC.
Storm? The last hours of the storm, tracked across both Washington DC and New York, missing New York state by probably inches!
Piers forecast; Public bulletin – WeatherAction 12 No 32 stated clearly 29th June to 1st July “Extended area of thunder + large hail, North Central to NE USA” showing the
track across the southern Great Lakes region. Confidence 75%. Timing normally to one day.
Storm? Severe thunderstorms with large hail across the region shown on the dates shown. Derecho subsided over NY at 4am 30th.
Several thunderstorms followed across the region over the next few days as per Piers 1st to 4th forecast.
I do note however, that I haven’t traced any reports of local floods, so you could add that as a fail I suppose.
From your own posted NOAA maps I see reports of large hail, eighteen on the 1st, one on the 2nd, none on the 3rd and five on the 4th.
How many reports of large hail do you need before you’ll believe that reality agrees with Piers forecast Willis?
I apologise for the added spacing; I suggest you blame Microsoft Notepad.
Russ says:
July 11, 2012 at 7:54 am
“I’m not letting you get away with that one. Large or Giant? By who’s definition? ”
Piers knows the answer to this one it seems.
He did indeed forecast ‘giant hail’ for the period 1-4 July.
He forecast ‘hail’ for 5-7 July
There’s a forecast of of ‘large hail’ for 13-15 July
We’re back to ‘giant hail’ on 16-19 July
‘Hail’ for 20-23 July
‘Large hail’ 24-28 July
Finally we have ‘massive hail’ for 29-31 July
So the Corbyn scale of hail has at least 4 sizes.
The importance of the work of Mr. Corbyn is, that he – as an astro-physicist – is trying tp translate the longterm in-and-outs of the astronomy (and specially of the sun and the moon) in terms of day-to-day meteorology.
Many other astro-physicists investigate the sun and its periods of changing sunspots and solar-flares and their influence on the climate on earth during centuries and decades. Mr Corbyn is trying to fill-in these decades in years, months and weeks for areas on the world.
Doing so, it is not important that there are differences with the reality of the day-to-day weather and it is logic that he afterwards picks up the many good predictions and ignores the bad ones. He now use method 8.A and I am sure that method 10.B will be better.
The huge importance of his work is that he explains climate change (and the weather from day to day) as a NATURAL process, influenced by the sun and the moon, and he ignores all that nonsense of ‘climatology’ of some particles CO2 more or less in the atmosphere.
Jaap de Vos says:
July 11, 2012 at 10:19 am
“The huge importance of his work is that he explains climate change (and the weather from day to day) as a NATURAL process, influenced by the sun and the moon, and he ignores all that nonsense of ‘climatology’ of some particles CO2 more or less in the atmosphere.”
Absolutely. You’ve hit the nail on the head. That’s why we should avoid cutting the grass from under his feet.
A forecast is a forecast, it is irrelevant what technique Mr Corbyn chooses to use.
Willis said “Maybe you see “distinct clustering” in that, but I see nothing of the sort.”
Fourier -he say;
Significant frequencies p<0.01:
12 to 22 years (the peaks ~13 years and ~22 years)
6 years
1 year
Agnostic says:
July 11, 2012 at 3:51 am
Surely Piers’s projections don’t depend on whether or not I have a car or whether it is hit by hail …
I looked for extreme thunderstorms in the predicted area during the predicted time. I found none. I looked for “giant hail” in the predicted area during the predicted time. I found none. I looked for “tornado swarms” in the predicted area during the predicted time. I found none.
Not sure how you turn that into a successful forecast, Agnostic.
First, if Piers is making predictions on a “bell curve”, then there is no way to falsify it. Even if it is a month out, the bell curve is not at zero, so he could claim success.
Second, Piers has said when these events are going to occur. You have never replied to my question about whether you count a Met Office forecast for a sunny clear weekend as a success if the weekend is rainy and stormy, but Monday and Tuesday are clear. Me, and everyone with a barbecue or a wedding on the weekend, will call that forecast a failure … and they will laugh in your face if you try to bullshit them with your nonsense about bell curves. A forecast is a forecast, made for a specific place and time.
I have no information, no data, and no documentation of any prior forecast. So I don’t have a clue a) if there was an earlier forecast, b) what it might have said, or c) how it differed from the current forecast.
Well, when you start claiming that everything is a success even if it is a week or two outside the forecast interval, I’d say there’s no way to tell if it was just a lucky guess.
Not only no tornados, but no giant hail and no extreme thunderstorms either … he missed on all three, and somehow you want to declare him “THE WINNAH AND STILL CHAMPEEN!” …
But no, I don’t need to say what I consider to be reasonable accuracy. That’s up to Piers. If he can only hit it within ± 1 week, as you seem to be saying, then he should widen the boundaries of his forecast so that they read “± 1 week”. And if he doesn’t do that, then you can’t count it as a success.
We don’t have a clue what his success rate is. If a cyclone in late July is a “success” when he predicted it in the first week of July, as you are strenuously arguing, he could be having a very high success rate.
In forecasting, a “success” is when what you say is going to happen actually happens when and where you forecast it. Anything else is a failure. If you forecast a nice weekend, then if it turns out sunny on Monday and rains on the weekend your forecast is a FAILURE. If you forecast floods in New York on Thursday and it floods in Boston on Thursday, that is a FAILURE. What I consider a successful prediction is what anyone (except evidently Piers and his followers) considers to be a successful prediction—the predicted event happens when and where you said it would, within the specified uncertainty bounds. Not the next week. Not in the next town over. When and where you said it would ± uncertainty.
Noted. For me, I don’t care when the forecast is made. Either it is right or wrong no matter when it was made. Certainly, if it is made a year in advance and is right, that’s more impressive that if it is made a week in advance. But no matter when the prediction is made, the rules are the same—if it happens when and where you said it was, it is a success, and anything else is a failure.
That’s a perfect example, if it actually happened like that. If so, then what Piers forecast came to pass where (England, or perhaps the UK) and when (winter 2010) he said it would happen.
But if he actually predicted a freezing bitter cold November and December and a mild January and February, and they happened the other way around, it would be a failure.
For a forecast to be valid we have to be able to call it a success or a failure. Your method, which seems to consist of “let’s give Piers the benefit of the doubt in all situations”, doesn’t allow that. I take Piers at his word. If he says a massive tornado will happen from the 9th to the 12th ± one day, then if it happens outside that interval it is a FAILURE as a forecast. If Piers wants me to consider a tornado on the 7th as a success, then he needs to forecast it from the 8th to the 12th ± one day, or from the 9th to the 12th ± 2 days. If he doesn’t, then his forecast is wrong. He made the rules, he set the dates, he specified the uncertainty … so he has to live with that and nothing else.
w.
Russ says:
July 11, 2012 at 7:54 am
No, he warned people to expect “giant hail” and “tornado swarms”. They got neither.
And predicting “thunder and hail” in the Midwest in the summer??? That happens almost every single day, that’s like predicting that water will be wet.
w.
Chas says:
July 11, 2012 at 1:01 pm
First, he had claimed that he could see “distinct clustering”, not reveal it through sophisticated mathematical methods. I was responding to his claim that there were visible clusters.
Second, I just ran a fourier analysis of the data and didn’t find the peaks you describe. Perhaps you could provide us with details of your data and your methods.
w.
July 4th storm report:
0640 PM TSTM WND GST 5 E STAPLETON 41.48N 100.42W
07/04/2012 E60 MPH LOGAN NE PUBLIC
DIME SIZE HAIL ALSO REPORTED.
0315 PM TSTM WND DMG SWANTON 44.92N 73.13W
07/04/2012 FRANKLIN VT TRAINED SPOTTER
LARGE TREE LIMB DOWN ON A POWER LINE
0418 PM HAIL 1 W WALDEN 44.45N 72.25W
07/04/2012 M1.00 INCH CALEDONIA VT PUBLIC
ESTIMATED 70 MPH WINDS BLEW A GRILL OFF A DECK AND
DOWNED MULT LIMBS AND POWER LINES… 1 INCH HAIL AS WELL
0422 PM HAIL 1 SSW WALDEN 44.44N 72.23W
07/04/2012 M1.00 INCH CALEDONIA VT TRAINED SPOTTER
POWER OUT AND 1 INCH HAIL REPORTED BY TRAINED SPOTTER
0424 PM TSTM WND DMG 1 WNW WEST DANVILLE 44.42N 72.22W
07/04/2012 CALEDONIA VT PUBLIC
15 TO 20 TREES DOWN ON CARS AND ACROSS N SHORE DRIVE.
SMALL BOATS FLIPPED IN JOES POND.
0430 PM HAIL WEST DANVILLE 44.41N 72.20W
07/04/2012 M0.88 INCH CALEDONIA VT PUBLIC
MULTIPLE TREES AND POWER LINES DOWN
0449 PM TSTM WND GST BARNET 44.30N 72.05W
07/04/2012 E60 MPH CALEDONIA VT TRAINED SPOTTER
TRAINED SPOTTER
0724 PM TSTM WND DMG 1 E SOUTH BURLINGTON 44.47N 73.15W
07/04/2012 CHITTENDEN VT TRAINED SPOTTER
UTILITY POLE FELL ONTO A SMALL AIRPLANE
0729 PM TSTM WND GST BURLINGTON 44.48N 73.21W
07/04/2012 E70 MPH CHITTENDEN VT TRAINED SPOTTER
KING STREET IN DOWNTOWN BURLINGTON FLOODED
0730 PM TSTM WND DMG 1 NNW BURLINGTON 44.48N 73.22W
07/04/2012 CHITTENDEN VT PUBLIC
MULTIPLE TREES DOWN ON N. CHAMPLAIN DRIVE
0734 PM TSTM WND DMG 1 SSE SOUTH BURLINGTON 44.46N 73.17W
07/04/2012 CHITTENDEN VT PUBLIC
TREES DOWN
0756 PM TSTM WND DMG SOUTH BURLINGTON 44.47N 73.17W
07/04/2012 CHITTENDEN VT BROADCAST MEDIA
FOX44 REPORTS A TREE DOWN ON I-189 IN SOUTH BURLINGTON
0803 PM FLOOD SOUTH BURLINGTON 44.47N 73.17W
07/04/2012 CHITTENDEN VT BROADCAST MEDIA
STREET FLOODING ON SOUTH WILLARD ST IN SOUTH BURLINGTON
http://forums.accuweather.com/index.php?s=&showtopic=29788&view=findpost&p=1566546
Willis “Perhaps you could provide us with details of your data and your methods.”
Using the .xls dataset from 1985;
http://floodobservatory.colorado.edu/Archives/
I turned the start dates into months using EOMONTH – produced a column monthly counts of events with a pivot table and then dropped the counts into ‘PAST’ (which is free software)
http://folk.uio.no/ohammer/past/
and ran a spectral analysis (Time -> Spectral analysis )
I then copied the frequency:power data from PAST into excel and turned the frequencies into periods [1/frequency*12] and selected the the periods that had a power of >11.1 .
Maybe too crude, as it contains the pattern of the number of days of each month?
What did you get?
In my – totally unprofessional – opinion, any forecaster who is willing to stick out his neck and go out on a limb to predict anything else than the standard national forecasts of “50% rain/50% sun” is worth at least some serious attention. Furthermore, nitpicking about a tornado being off one day probably is less hazardous than NOT predicting a tornado, at least for those directly concerned. At least people are being made aware of the chance of a touchdown, right? If I were a professional climatologist/weather person, and I was confronted with somebody who manages to at least score a very good percentage of indicating when certain weather types are most likely to strike, I for one would probably really like to know which method was used for these percentages, and explore it further. Why all the status quo quarrels? Why not join forces and see if everybody involved can come up with a combined system that finally guarantees (to a certain extent) that folks like me – who only know which way the wind is blowing by simply sticking our finger in our mouth and holding it up – can be told whether we can get out out bathing suits or need to shore up the house because of floods? I live in an agricultural area. Farmers don’t give a hoot about a deluge coming on either the 22nd or the 23rd. What they do care about is knowing if the month of July is going to be wetter than usual. Once rain is coming, they are perfectly capable of “feeling it in the air” and by watching how nature behaves around them. They care about being told whether winter is going to be hard and long (no pun intended) or short and sweet. They care about being able to take into account whether the potato harvest is going to be normal this year, or whether they need to financially take into account that their crops will be slow this year. You see, it’s not about ‘who’s right and who’s wrong’. It’s about direct consequences for the public in general. So gentlemen, please get off your high horses and get down to basic facts. Share information. Who knows, you might even learn something from each other 🙂
I’ve been a subscriber to Weather Action forecasts for years. His ability to predict extreme weather events is almost astounding. Admittedly, and even he will admit that he does not get it right all of the time, but definitely more so than the Met Office who actually gave up long range forecasting because they were so bad at it.
In the past, when the MO have predicted their ”BBQ global-warming summers”, Piers’ has belly-laughed at them and then even more so when MO are proven wrong. The same goes for the last few bad winters when MO were predicting ”our children will not know what snow is’. If is great fun to be a witness of this too. You could also look at the older winter predictions for USA, the tornado season, heat waves and other weather events.
Many people think that they are badly odd, but the ‘thing’ with Piers’ forecasts is learning how to read them. Sometimes they do seem to go out of kilter, but looking at the weather fronts on a standard met map you can estimate for example if nothing seems to have come right, then you can determine if the events are still going to come, or where they are if they missed the predicted area.
We use ours on a ‘monthly weather trend’ basis, but we do monitor the individual periods and make plans accordingly, especially in Winter where things can become very problematic overnight.
To see clustering clearly your need tuned windowed analysis.
Not sure how you turn that into a successful forecast, Agnostic.
By looking at whether that type of wether event occurred close to the period.
In forecasting, a “success” is when what you say is going to happen actually happens when and where you forecast it. Anything else is a failure.
Wrong.
You can say it is a failure if you are talking about whether the weather is sunny or or cloudy a week in advance, but if you are predicting severe weather a month in advance, and you are able to do that more than 50% of the time then you clearly can detect something standard meteorology cannot. I don’t know why you can’t see this. Your position it is utterly unreasonable in my opinion, and it is a great shame. You are a well respected climate science auditor and when you put your mind to something you turn out really interesting analyses. But you seem to be to willing dismiss the SLAT methodology in an imprecise science presumably because Piers presentation doesn’t conform to your sense of scientific impartiality.
You seriously throwing the baby with the bath water here.
First, if Piers is making predictions on a “bell curve”, then there is no way to falsify it. Even if it is a month out, the bell curve is not at zero, so he could claim success.
Rubbish. But at least you have started to narrow the time frame – clearly a month out is too far out for the prediction to be useful. But if it misses by a few days it’s not….?
You have never replied to my question about whether you count a Met Office forecast for a sunny clear weekend as a success if the weekend is rainy and stormy, but Monday and Tuesday are clear. Me, and everyone with a barbecue or a wedding on the weekend, will call that forecast a failure … and they will laugh in your face if you try to bullshit them with your nonsense about bell curves. A forecast is a forecast, made for a specific place and time.
This is where it is clearest you are missing the point of what Piers is doing (or doing apparently because at this stage I am not sure). My reply to the question about the MetOffice, is that they often predict sunny weekends that turn out to be stormy (or get a near time forecast wrong) but I don’t then dismiss every subsequent forecast they make! Further more they have a very poor record at long range forecasting which is the point of interest in this discussion!
The SLAT thing is about long range forecasting. On the BBC website they explicitly say “The weather beyond about a week ahead stretches even the most experienced weather forecaster.” Yet Piers – apparently – is able to forecast accurately from longer than that. That is what the whole fuss is about. And by accurately, the further back you are forecasting the greater the range of error you have to allow surely…? But then you say:
For me, I don’t care when the forecast is made. Either it is right or wrong no matter when it was made.
Whaaat? So supposing someone makes a prediction that there was going to be a severe storm that rains frogs 1 year before it happens and it occurs a few days later or earlier and rains tadpoles instead – that is going to be of no consequence or interest to you? And they were able to make those kinds of predictions 85% of the time? That is meaningless is it?
From a societal point of view, if this SLAT method is able to detect (eg) flooding events 1 month ahead of time and be able to say roughly when and where, I think that is extremely important! But you are caught up in the ‘…but he said…’ nonsense. Maybe he is exaggerating his claims, maybe he is not and this is a rare miss. Maybe there are extenuating circumstances. I don’t know and neither do you. I was just really hoping you’d find out.
That’s a perfect example, if it actually happened like that. If so, then what Piers forecast came to pass where (England, or perhaps the UK) and when (winter 2010) he said it would happen.
It did happen! Here is Boris Johnson, the Mayor of London’s take on it:
http://www.telegraph.co.uk/comment/columnists/borisjohnson/8213058/The-man-who-repeatedly-beats-the-Met-Office-at-its-own-game.html
FWIW his position is pretty much the same as mine here. I want to know more – is he on to something or is he just a fluke artist? But your approach is NOT going to find that out.
But if he actually predicted a freezing bitter cold November and December and a mild January and February, and they happened the other way around, it would be a failure.
Yep and fair enough. And he did predict a cold January and February this year which did not eventuate. He got the December cold snap right, said it would turn milder and then much colder at the end of Jan, but it stayed relatively mild. So he does not always get it right. He even, cagily, admits the errors. But understandably he does not advertise them too much because he has an interest in emphasising the successes – it’s his business.
For a forecast to be valid we have to be able to call it a success or a failure. Your method, which seems to consist of “let’s give Piers the benefit of the doubt in all situations”, doesn’t allow that. I take Piers at his word.
So much wrong here. For a long range forecast we need to be able to say whether a severe whether event occurred close enough to the predicted time (what constitutes close enough? I don’t know) close enough to the predicted area, or that the general weather conidtions (for example the very cold May in the UK) actually come to pass. It’s an imprecise science and your pleading for precision is NOT reasonable. My method is NOT to give Piers the benefit of the doubt and it’s ridiculous to suggest that – my method is to try to determine whether or not the SLAT method has any ability to detect severe weather that could impact on society or whether it can detect conditions reliably such as the 2010 big freeze – which it apparently did.
And if you were taking Piers at his word, then you should read again what he wrote up thread regarding worrying about the jockey’s socks when the horse he picked won the race. But I don’t think you SHOULD take him at his word – that is the whole point. He has a vested interest in establishing his method as a success. It needs independent and objective evaluation. I’d have thought you would be ideal at doing this – but maybe not. It probably needs someone who doesn’t mind getting a bit down and dirty with probabilities and statistics. It may not be ‘cut and dried’ enough for you – no shame in that.
Is Weather Action actually a business or a hobby that makes a little money? All this talk of ‘it must be successful because he’s still in business’, yet I can’t find it listed anywhere.
Agnostic says:
July 12, 2012 at 8:10 am
So if on Wednesday the Met Office says the weekend will be gorgeous, and you plan your barbecue, and it is fine on Thursday and Friday but it rains heavily all weekend with giant hail and tornado swarms, that’s a success because that type of weather event occurred close to the period … I’ll keep that in mind next time you start telling me that the Met Office is a failure.
Do you realize how crazy that is, to call a forecast “successful”, not because it was a success, but because you missed it by an unspecified amount?
Hey, if every time I get it close you count it as a success I can be right more than 50% of the time, that’s easy. Here’s why.
If I simply predict what has happened in the past, the nature of averages says I’ll be right a good chunk of the time. But if my forecast is counted as being right when it’s only close, I can be right much more of the time.
Here’s an example. Suppose it rains one day out of three on average. Suppose further that I always forecast clear bright sunny days. Since the odds are quite large that there will be a clear bright sunny day within ± 2 days of the rain, and you say that we determine my success by “whether that type of weather event occurred close to the period”, virtually every one of my forecasts will be a “success” … as you define success.
I’m not “dismissing the SLAT methodology. As I have pointed out numerous times, I don’t have anywhere near enough evidence to dismiss anything. I’m simply pointing out that when a man says there will be “tornado swarms” in a certain area during a certain time span, and not only are there no “tornado swarms”, there’s not one tornado in the whole region during the whole time, he’s wrong and his forecast was a failure. How tough is that to understand?
The question is not, and has never been, whether Piers’ forecasts are useful. Heck, if he were wrong 100% of the time they would be very useful, just bet the other way. The question is not whether they are useful but whether they are correct, and how often.
My question was whether you would consider the Met Office forecast a failure. You have expended a lot of words in not answering the question. Is such a forecast a success because that “type of weather event occurred close to the period”, or not?
I’m not “dismissing every subsequent forecast” of Piers, what on earth gives you that impression.
I agree. The further back in time Piers is forecasting, the greater the range of error that PIERS has to allow. But if he doesn’t allow a greater range of error, if he specifies it to within three days, then that’s what I’ll judge him on.
Come back and ask me when Piers can predict a rain of frogs one year out 85% of the time. Until then, that is idle speculation.
I’m trying to find out. The only way to find out is to look at the accuracy of his forecasts, using some kind of criteria. Your criteria seems to be “well, it was kinda close, it’s a success”. This is science, Agnostic. The concept of science might be easier for you if you think of it as a bet.
If I bet you the weekend will be sunny, and it’s pissing down rain, who wins the bet?
You do. Duh.
Can I claim that I won the bet because Friday and Monday were sunny?
No way, that wasn’t what we bet on.
Do you see how this “science” thing works now, Agnostic? Either what you bet on is right or it is wrong, and “close” doesn’t count.
Agnostic, perhaps you think the Mayor of London is some holy referee to settle the question. For me, the only way to settle the question would be for Piers to make public his past forecasts, so we can determine if they were right. Not ask the Mayor of London if they were right. Determine scientifically and objectively if they were right. I’d like nothing more than to do that … but a) Piers hasn’t released the forecasts, and b) Piers, you, and his followers claim that every forecast thats sorta kinda right is a resounding success.
But wait, you said if the event occurred close to when he predicted it, it was a success. If he predicts “cold December, warm January” and the reverse happens, by your terms both forecasts should be counted as a success. Now, you say they are a failure … please make up your mind.
Yeah, that’s what I said, your method is ‘lets count failures as successes if they are close’.
While the science is certainly imprecise, the outcomes are not—either the weekend is sunny, or it rains. That’s the part you seem to be missing.
Oh, please. It is obvious that I meant that when Piers is making a forecast I take him at his word as to the boundaries and nature of the forecast. If he wants to talk about the meaning of life or the color of his jockey shorts, or if he wants to claim his forecast is a success, no, I don’t take him at his word for any of that.
Oh, enough with the insults, it makes you look spiteful.
As I have said before, Piers is setting the boundaries on his forecasts, and I take him at his word about those boundaries. If he says there will be floods in California, I assume he is not talking about Nevada. If he says the floods will happen the 4th to the 7th of June, I assume he is not talking about the last week in May. He makes the forecasts, and he sets the parameters. If he wants Nevada included, he needs to say so. If he wants the last week of May included, he needs to specify that. I am merely trying to see if the predicted outcomes fit within the boundaries that he himself has set.
But then you come along and say hey, Arizona is close to California. And hey, the last week of May is close to the first week in June … so his forecast of floods in California in June is a resounding success because it flooded in Phoenix, Arizona in May! Another amazing win for the SLAT method!!
I’m sorry, but that is not science, that way lies madness. If he meant Arizona, he should have said so. If he meant the last week of May, he should have said so. The only way to objectively analyze his forecasts is to see if they were right or wrong under his specified terms of the forecast, not to see if something similar happened in Phoenix in May.
w.
PS—Let me remind readers again that after Piers asked if anyone wanted to bet on rain for the Olympic Opening Ceremonies, I said I was interested and asked him what the odds and the terms of the bet were. Since then … just the sound of crickets.
Willis, there was an independent study done. The results of it are on Piers’ website here:
http://www.weatheraction.com/pages/pv.asp?p=wact45