## Discussion > Predictions for 2016

EM, we're only 3 or 4 years away from the point where the pause becomes statistically (95%) likely. A the moment, the trend is consistent with this projection. You may choose to say it's not there until it hits the rather arbitrary 95%, but if I was gambling on a coin toss, I'd be happy to go with odds much less than 1 in 20. If it gets to the end of 2016 (as I said in my original post) then the probability of the pause being real is already 67%, already enough for most sensible people to begin acting as if it was a real possibility.

Martin, yes thermometers use a proxy; the expansion of a metal or liquid, the change in resistance, the spectrum of IR. Some are directly related to the object temperature, sometimes with some processing (e.g. IR thermometers). A satellite measuring emissions of oxygen in the columns of air beneath and to the side of it is not really the same. If what was done was to measure the emissions from directly below the sat and call that the point temperature of that column of air, it would be comparable to, say, an IR thermometer. But that isn't what is done as there needs to be some secret sauce in which the cooling of the stratosphere is removed (cooling not warming) and there's some assumptions about standard atmospheres, which don't exist, and there's adjustments to remove high latitudes and other parts of the signal where the signal is not usable, and, and, and.... Like I said, I'm sure you are happy with describing it as "measuring" temperature, just like you would say a telescope measures temperature. I guess I will never get used to "skeptic" language - but anyway I modified my no. 9 to include the word "tropospheric" - satellites may "measure" the bulk temperature of what is beneath them, but they surely don't measure the internal temperature (just like a thermometer stuck in my mouth doesn't measure my core temp - although it can be used to infer something about it).

Whether some people might think that temperature measurements being a weighted average over a range of altitudes is an advantage would depend on what they were trying to measure. If the troposphere is warming and the stratosphere is cooling, a weighted average of the two is largely useless if they want to know what is happening near the surface - but maybe they don't.

Another thing I guess I'll never get used to is how "skeptics" can talk about a continued pause as if the record levels of this year and last don't exist. You and Big Yin seem to enjoy this fantasy and even supply graphs to WFT to "prove" your point with the data (click "Raw Data") going all the way up to 2014.33 (0.534 C) or perhaps 2014.42. Way to go, "skeptics"!

All the same I bear you no ill will. Happy Christmas and New Year!

@EM Temperature graphs don't shadow the CO2 graphs these past years.

In the nineties we were told by the authorities that "see how these 2 graphs follow each other".

..Well now they don't people are skeptical.

You can that on the Graph Fit on the Wood For the Trees link above

.....Merry Christmas !

TheBigYinJames

May I ask how you calculated these probabilities? What were your start and finish years and your method? Do you have the data available for me to check the calculation?

Mine come from the confidence limits(+/-2SD) of the GISS data, shown on their annual average graphic as +/-0.09C.

Statistics 101 gives the threshold for a 95% probable significant difference between two sample means as four standard deviations . For GISS data this is 0.18C.

If I cherrypicking the sceptics' favourite year of 1988 and compare it with 2015 the difference is 0.84-0.63=0.21C. That is statistically significant.

If I choose 1997 the difference is 0.84-0.48=0.36C. That is also statistically significant.

If I choose 1999 the difference is 0.84-0.42=0.44C Once again statistically significant.

If I use the 5-year average centred on 1998 and the most recent available 2012 the difference is 0.67-0.46=0.21C. Significant again.

If I repeat this exercise with any of the surface temperature datasets you get a similar result. There has been statistically significant warming over the last two decades.

Radical Rodent

. As you say, choose the starting point which is most sensible. Which are your preferred years for the start of the long term trend and the pause, and why?

Simple, really. My start point for looking backwards is now. I then look to see how the trend has been; the point where the trend becomes noticeably different becomes is then “cherry-picked”, to become the next reference point. Thus, the time when I parked my metaphorical car was about 1998, having started the journey around 1975, following a shorter, slower journey the other way, from 1945.

To use another analogy – at what point do the various plateaux of Ingleborough start or stop? Or are you “cherry-picking” those places where the path no longer rises for many paces, and it is just a single continuous slope from the base to its summit, if you overlook the inconvenient flat bits? Does the fact that the many points on its plateaux which are higher than others prove that they are not really plateaux?

Hey, EM, why have you not critiqued **my ** 'elementary data analysis'?

I doubt that it contains more implicit assumptions than your own - perhaps fewer.

As I've said before, if you can't validate your assumptions, or you don't even know what they are - such as what is the stochastic process generating the time series, then statistical analysis is just guessing in fancy dress.

Martin A

I was just looking at your graph.

I took the liberty of adding the long term trend. If I saw that data in a neutral context I would regard it as indicating an upward excursion in the late 1990s followed by a regression to the long term trend.

It will be fascinating to see which line the data follows when the two trends converge. If it stays flat you win, if it reverts to the long term trend, I win.☺

MartinA

*If it stays flat you win, if it reverts to the long term trend, I win.☺* And if it goes down?

I'd certainly agree that it's a guess. But I don't see it as a bet like that.

Somebody stating that currently we are in a pause is not saying what might happen in the future. They are not saying it will *remain* flat. They are not saying it will revert to the "long term" (whatever that might be) trend. They are just saying that, currently, it does not seem to be going up neither does it seem to be going down.

Personally I prefer the term "halt". Becasue "pause" implies that we know that it's only a matter of time before it starts rising again. Which we don't

I see "halt" as neutral - it does not imply that it will stay constant forever (any more than "Halt at Major Road Ahead" did not used to mean you'd stop there for more than a short time). But neither does it imply that it will soon be rising once again.

Hey, EM, why have you not critiqued my 'elementary data analysis'?Perhaps because I already did, above, pointing out that your data stops in early 2014. As if cherry picking a start date isn't enough to invalidate any significance your "analysis" might have.

Radical Rodent

Climbing Ingleborough was interesting, especially when the gradient increased on the grit.

I confess that I cheated. I set up camp in a field behind a pub and did the three peaks over three days.

thaks raff. all i did was fit a line and display it. i did not say what if anything it signified. my view is that it signifies nothing bcos the underlying processes are not undersood. but since the slope of the line is ~zero 1998 - 2016 it seems to me to show that in one sense the pause is undeniable. that says nothing about its significance if any.

i wanted em to question my assumptions

Well one assumption was that if you put 2016 into WFT you'd get data up until the present, when in fact you got data up until early 2014.

but since the slope of the line is ~zero 1998 - 2016 it seems to me to show that in one sense the pause is undeniable.no that slope ends in 2014, as I've now said several times. Look at the raw data if you don't believe me:

...

2014.08 0.27

2014.17 0.488

2014.25 0.578

2014.33 0.534

#Data ends

#Number of samples: 893

#Mean: 0.0645666

...

Another assumption is that cherry-picking a start date is a valid way to determine a meaningful "trend".

*no that slope ends in 2014, as I've now said several times.*

Raff, what you say looks right. How do you make WFT list the data values, rather than graph them?

*Another assumption is that cherry-picking a start date is a valid way to determine a meaningful "trend".*

I don't think I made any claim that the straight line from 1998 to ~~2016~~ 2014 determined a *meaningful* trend. Others might well do, but not me. I just computed a line from a date that I liked the look of and plotted it. I don't think I said anything about what the line implied.

The following might explain why you see things differently from me.

**viewpoint A; It's just a line**

I just regard it as a line, computed using a standard formula. It's just a line. Only when someone starts explaining what is the significance of the line do they have to start discussing whether it is *meaningful*

But it's undeniable that the slope of a least squares regression straight line provides a measure of the average rate of change of the *actual* data. To repeat, it describes a characteristic of the actual data. It's not open to doubt (apart from accidental errors) any more than "the sum of the actual N data points is X1 + X2 + ... +XN" is open to doubt.

So I don't understand how anyone can deny that the rate of change of actual temperature has averaged to about zero since 1998. In that particular sense, I think that the existence of 'the pause' is undeniable.

**viewpoint B; It tells us about the underlying random process**

The slope of the regression line can also be *interpreted* as providing an *estimate* of one of the characterisitcs of the underlying random process that generated the data. As soon as you make that interpretation (as I think EM does) you immediately make a whole load of assumptions; For example some frquently made assumptions seem to be:

- There is a knowable (in principle) underlying random process

- It generates values consisting of a linear trend with random fluctuations added

- I think EM also assumes implicitly that the random fluctuations are stationary (their statistics don't change over time), Gaussian, and uncorrelated (which, if they are Gaussian, also means they are independent)

If it is providing a reliable estimate of the characterisitcs of the underlying random process then clearly it should not be sensitive to the fine details of starting point and so on. So I understand the objection to choosing 1998 in particular if sombody is using it to infer things about the underlying random process.

**Bottom line**

I think people who say "the pause is a myth" (or something similar) mean that they see no reason to think that the underlying random process has suddenly changed its nature merely because of what the recent data points happen to be.

I think some people (like me) who say "the pause/halt is undeniable" mean simply that the actual temperature data has not shown much of a change over the last 18 or so years.

Does that make any sense?

Of course some people (also like me) think that the pause/halt shows that the climate system is understood far less than climate science had thought but I reach that conclusion for a load of reasons that have nothing to do with regression analysis:

To get the data, click on "Raw Data" under the graph (as I said before - I get the feeling you don't read my laboriously crafted prose :-( ) where it says "Audio waveform - Raw data - Charity Tip Jar".

So I don't understand how anyone can deny that the rate of change of actual temperature has averaged to about zero since 1998. In that particular sense, I think that the existence of 'the pause' is undeniable.Well from the graph you presented nobody can deny that rate of change of the "HADCRUT3 variance adjusted global mean" index was about zero between 1998.0 and 2014. I found it was in fact

**exactly zero**between 1998.0 and 2013.6 (enter the decimals in WFT) judging by the lack of aliasing. Do I get a prize (some cherries perhaps)? You too are making some big assumptions if you thing that this zero applies to "actual temperature". Try another index with the same dates and you'll get another value.

As to whether people think there is a knowable underlying random process, that sounds like a contradiction. If it is random, it is not knowable, by definition. And I doubt very much that people who draw lines from 1998 are interested in understanding any underlying non-random process that might exist; quite the reverse. As a Christmas exercise, you could use decimal dates (2013.6 etc) to try finding an exactly zero least squares regression somewhere between 1998 and now on the other indices. I'm sure you'll find some, perhaps many :-)

Happy Christmas and New Year.

Raff - sorry not to have given your carefully crafted prose the attention it warranted. That is the general fate of carefully crafted blog comments - to be skimmed rapidly and maybe get a response dashed off before the cup of coffee is empty and other things call.

*You too are making some big assumptions if you thing that this zero applies to "actual temperature". *

I don't think that. I don't even think there *is* such a thing as "actual temperature" (because, among other reasons, there are an infinity of ways of defining average global temperature, none of which has any better claim than the others to representing something meaningful).

*As to whether people think there is a knowable underlying random process, that sounds like a contradiction. If it is random, it is not knowable, by definition.*

No it's not a contradiction. The random process is the algorithm or the mechanism that generates the sequence of random values. The process itself in principle can be known, even though the random values it generates cannot be predicted in advance. It can contain non-random components. But if the physical system is not properly understood, then it cannot be analysed for the random process to be known.

For example here are two random processes generating random numbers at t = ...-2, -1, 0, 1, 2, 3, .... The number at t = k is denoted by xk. The {rk} are independent uniformly distributed random numbers with zero mean and standard deviation 0.083.

Random process A:

xk = 0.2 × t + rk

[ie xk contains a non-random component proportional to t, (a linear ramp) with uncorrelated random values added] The long term trend in A is 0.2 per unit time.

Random process B:

xk = 0.93 × xk-1 + rk

[the {xk} are a sequence of random numbers of zero mean and with strong correlation between successive numbers] The long term trend in B is zero..

Here are a few samples from each random process, carefully selected to look a bit similar.

t _ _ A _ _ B

1 0.44 0.69

2 0.38 0.93

3 0.78 1.34

4 1.11 0.86

5 0.90 1.26

6 1.67 1.59

7 1.57 1.43

8 1.36 1.64

9 1.82 1.77

10 2.44 1.68

11 1.78 1.68

12 2.55 2.03

13 2.48 2.23

14 2.51 2.20

15 2.68 2.52

16 2.80 2.36

17 3.14 2.57

18 3.80 2.80

19 3.39 3.04

20 3.94 3.05

The only point of showing these numbers is to illustrate what is meant by a random process. Plus also perhaps to illustrate that quite different random processes can generate similar looking sequences over a shortish time interval.

Many thanks for your good wishes.

I don't think that. I don't even think there *is* such a thing as "actual temperature"Yes I understood that from previous discussions with you. But above you did say

So I don't understand how anyone can deny that the rate of change of actual temperature has averaged to about zero since 1998.implying that you think you can infer something about "actual temperature" from the slope in your graph. I don't see how you can have it both ways.

I take your point about random processes (about which I clearly know nothing).

Predictions:

1. We will vote to leave the EU (how Cameron responds to that is another matter).

2. Cameron will quit or otherwise lose his position as leader of the Conservative Party.

3. Corbyn will lose the leadership of the Labour Party.

4. Cruz will be nominated as the Republican Presidential leadership candidate and he will become the next president.

Well I think I stuck my head into enough nooses for one day now hehe.

So we're moving back on thread? I have a couple more

1. Raff will continue to divert threads from the original topic until he goes for a two week holiday in August, when he gets back he'll have found another diversion.

2. Some small group of greens will discover that Steel contains carbon and campaign for Carbon Free/low carbon Steel. As for carbon fibre bicycles they'll be treated as worse than Chelsea Tractors.

SandyS: point 1: a diversion is only a diversion if you follow it. It might be better for you, and many others, to follow the advice of Mike Jackson, and ignore the little twerp, whose arguments contain little in the way of rationale or logic, and who refuses to answer any direct question, like so many others who *claim* to be offering counter-arguments, but who are really engaged in trolling. Ignore it – DNFTFT!

Point 2: the Chelsea Tractor is actually the transport of choice for most “Greens”, as they feel it makes them look to be in touch with the soil, *á la* those servants of toil, the farming community (that they are actually contrary to practically all that the farming community are involved in is another dichotomy that they can bizarrely live with). (Also, driving a Chelsea Tractor absolves you from most of the highway code, as you can drive and park wherever you like, and all other traffic must give way to you.)

A couple of my own predictions:

1. Whatever the Met Office predicts for summer, the weather will do the opposite – I suspect that they will predict a wet, miserable year, so look forward to a balmy few months – with perhaps even a drought being declared (and blamed on… guess what?)

2. Some “emergency” will occur, causing the US presidential elections to be cancelled, and Obama remaining in office, with increased powers to over-rule both Houses, until it is resolved – if, indeed, they ever think it can be.

To the Rolls Royce of rodents:

I think your second point is an enormously positive but still wishful bit of thinking hehe.

Ted Cruz will not be denied.

At some point in the next five years (maybe 2016?) the pause will end!

A period of cooling will begin which will end in cold place ^.^

It is like my car that has been sitting on the driveway, stationary (apart from rocking in the breeze), now, since 9 o’clock last night. Why did I cherry-pick 9 o’clock to declare it being stationary? To me it seemed a good time to pick, as that was when I stopped there; there are many, like EM, who might say that I could well have chosen 8 o’clock, when I was 50 miles away, so proving that my car is not stationary, but still travelling at a little over 3 miles an hour. Hey! Maybe I was wrong to cherry-pick the times 8 o’clock to 9 o’clock, and claim that I had averaged 50 mph; had I chosen 7 o’clock, when I was 80 miles away, my (cherry-picked) average speed would have been 40 mph, but really would continue to be about 5 miles an hour. Oh, dear… what point should I choose to avoid being accused of cherry-picking?

And, how does this affect any predictions for 2016?