Buy

Books
Click images for more details

Support

 

Twitter
Recent comments
Recent posts
Currently discussing
Links

A few sites I've stumbled across recently....

Powered by Squarespace

Discussion > Hypothesis testing in climatology

This started on Unthreaded on November 9th.

Michael hart rejects the validity of any science not done using controlled experiments which excludes all observational science. Geronimo says that hypothesis testing and falsification cannot be done in climatology for similar reasons.

I contend that they are mistaken.

Firstly we need to establish what we agree on.

Michael hart, geronimo,; can you indicate which of the following do you accept as valid evidence. There is no point doing this if you do not accept measurements taken outside the laboratory.

IR absorbtion and emission spectra of CO2, water methane and atmospheric air measured under laboratory conditions.

IR absorbtion and emission spectra and intensities of outward longwave radiation by satellite and downwelling longwave radiation measured from the surface.

Global average temperature measurements averaged from stations, satellite measurements, ship measurements etc.

Measurements of deep ocean temperature by ARGO and earlier techniques.

Laboratory measurement of coefficients of expansion, heat content and latent heat for seawater and freshwater.

Measurement of sea level rise using satellites and tide guages.

Measurement of volumes and volume losses of ice sheets by survey, satellite surface profiling and gravimetry.

Trenberth's energy budget and updates.

Nov 9, 2014 at 12:47 PM | Unregistered CommenterEntropic man

Jamesp

What I propose is to derive the increase in energy content of the land, ocean atmosphere from the CO2 hypothesis and show that it matches the measurements, using two independent approaches.

It will take a little while to assemble all the bits of paper, bear with me.

Nov 9, 2014 at 1:16 PM | Unregistered CommenterEntropic man

EM, this is what I had largely written before you said you were going to start a discussion thread.

Michael hart

You insistance that the only valid science is done by controlled experiments eliminates all observational science. For example, the only fact you can accept from astronomy is that the Earth rotates. ( Foucault pendulum )..

Nov 9, 2014 at 10:18 AM | Unregistered CommenterEntropic man

EM, the “facts” that you choose to accept will depend on the standards you choose to accept.

I deliberately chose my words to avoid slurring other scientific disciplines, but you've managed to interpret them the wrong way. Other fields have their own specific issues, and judgements can be made about how the practitioners take steps to mitigate those problems. For example, the search for the Higgs Boson reportedly required 'five standard deviations quality' data before it became publishable. Much other published research only requires a lower 'two SDs' (the "95% confidence level") before it is considered 'acceptable'.

These are, of course, completely arbitrary choices within the respective disciplines. Arguments to tighten-up such standards can be made, and are.

Now, let us look at IPCC 'climate-science': When the IPCC model ensemble failed their own '95% confidence level[*]', what did they do? They went in the other direction! They downgraded the pass level to 90% and then claimed an increased level of certainty! By continuously lowering standards they could make the models 'correct' from the beginning of the solar system until the end of eternity, if they wanted to or were allowed.

I put it to you that those are not standards that pertain in most of the physical sciences. Some might say that lowering standards in the face of hypothesis-breaking results is tantamount to having no standards at all.


[*As RGB at Duke frequently points out at WUWT, the IPCC model ensemble mean and associated confidence levels is a scientific abomination, but here I'll take it as a genuine attempt to convey uncertainty to an audience]

Nov 9, 2014 at 5:46 PM | Unregistered Commentermichael hart

Michael hart

Feel free to go elsewhere. Your denial of the validity of the measurements I intend to use means that anything I say here will be irrelevant to you.

Nov 9, 2014 at 6:19 PM | Unregistered CommenterEntropic man

I'm not michael hart nor geronimo but, off the top of my head, here is my assessment, as a casual onlooker, of my confidence in the (accuracy) × (reliability) of of the things in that list:

IR absorbtion and emission spectra of CO2, water methane and atmospheric air measured under laboratory conditions.
95%

IR absorbtion and emission spectra and intensities of outward longwave radiation by satellite and downwelling longwave radiation measured from the surface.
75%

Global average temperature measurements averaged from stations, satellite measurements, ship measurements etc.
27% [perhaps I'm being too charitable here in view of the inherent nonsense in the idea that temperatures can averaged, the coming and going of measuring stations over time and their haphazard spatial distribution]

Measurements of deep ocean temperature by ARGO and earlier techniques.
30%

Laboratory measurement of coefficients of expansion, heat content and latent heat for seawater and freshwater.
95% [heat content? (= specific heat?). For "seawater", how representative will be the sample(s)? ]

Measurement of sea level rise using satellites and tide guages.
55%

Measurement of volumes and volume losses of ice sheets by survey, satellite surface profiling and gravimetry.
45%

Trenberth's energy budget and updates.
10%

Nov 9, 2014 at 6:26 PM | Registered CommenterMartin A

EM: "If you can't explain your theory to a barmaid, it probably isn't good physics." Kelvin.

Let's get on the same page shall we, I accept that CO2 is a GHG and that there could be AGW. Without catastrophic bit it would be interesting but not worth arguing about. So here's what I think the theory your talking about is:

CO2 is a GHG, it has increased in the atmosphere because of human emissions, and temperatures have risen. Ergo CO2 causes temperatures to rise (or it used to). The rise in temperatures will cause untold catastrophes.

I'm not asking you to prove anything to me, we've spent $100s bn on climate science research and you are unable to point me to the simple tests that would prove the theory put forward by the scientific community. Moreover when a theory is proposed the criteria that would prove it wrong should be stated, so we don't have to argue it here, it must already be in the scientific literature if it's a proper scientific theory. But what falsifies the above theory as expounded by me?

As it happens I believe that CO2 causes an increase in the heat retained in the atmosphere. Whether this extra heat will manifest itself in the climate system is more complicated, and I'm pleased to see that the entire climate scientific community are as ignorant as I am, given the 55 (?) explanations for the pause.

What I don't believe is that anyone is equipped to foretell the future of a complex system that is interacting with at least two other chaotic systems.

Nov 9, 2014 at 7:07 PM | Unregistered Commentergeronimo

JamesP

Let the fun begin. I haven't tried quite this path before, so it will be fun to see how it works out. :-)

Prediction.

The physics of AGW theory predicts a radiative forcing imbalance due to GHGs of 2.706W/M2 for 2007, increasing to 2.916 for 2013 (Radiative forcing table, Wikipedia). Average for those seven years comes out at 2.8WW/M2.

(I chose 2007 as a start point because this is the year ARGOS became fully operational. More on this later. )

How does prediction compare with observation?

The observed aerosol forcing is between -0.3 and -1.8 ( IPCC2007) .Put them together and one would expect a maximum forcing of 2.8-0.3= 2.5. The minimum would be 2.8-1.8 =1.0. (Other positive and negative forcings such as cloud also play a part, but wwe'll keep it simple.)

Satellite measurement of insolation and outward radiation give an imbalance of 0.7+/-0.3 (Murphy 2009) to 90% confidence.

For this first test the maximum probable observed value and the minimum prediction meet. Not bad for a quick calculation.

Next I'll start playing with energies.

Nov 9, 2014 at 8:33 PM | Unregistered CommenterEntropic man

Martin A, Geronimo

As I said to michael hart, if you don't believe the measurements, this thread won't be relevant to you. I take the data and their confidence limits at face value and calculate accordingly. They tell a coherent story, which ,I would not expect from bad data.

Nov 9, 2014 at 8:39 PM | Unregistered CommenterEntropic man

Martin, do climate scientific publications claim that the values they calculate or measure are 100% accurate? Or do they give likely error margins? How do your % numbers relate to these error margins. If they don't, what do they mean? How are they derived and what do they really signify. Do they have error margins themselves? I mean 27% is awfully specific for something so nebulous as your degree of confidence in global average temperature measurements. How did you come to 27 and not 26 or 28? These look like numbers plucked from the air. And only 95% for lab measurements? Does that mean that there's a 1 in 20 chance that any of them might be out by a smidgen or out by a significant amount? I'd have thought there was a near 100% chance that they might be very slightly wrong to a minuscule degree. But what of it?

Nov 9, 2014 at 8:45 PM | Unregistered CommenterRaff

Raff, you are looking for things that are not there.

These look like numbers plucked from the air.

That's exactly right. As I said "off the top of my head, here is my assessment".

I'm not sure why you think 27% seems any more awfully specific than any of my other guesses, for example, 30% or 45%. They are all given to two figures.

As I tried to explain, I was giving a figure for my off-the-cuff guess about both the accuracy and the reliability of the things that EM mentioned. I didn't define what I meant by accuracy and reliability and I suppose that they can't really be considered as separate things.

But something that has been measured to a precision of ±10% and where you can have complete confidence in how the measurement was conducted, and the reliability of the reporting of the value, would show up as 90% on my subjective estimate.

Whereas something that has been measured with a precision of ±5% but where problems in gathering the samples, recording the results, using the wrong formula and so on meant that the result might (50:50) be very far from the true value might show up as "45%" on my subjective scale.

I hope that makes sense, even if you don't agree with my guesses.

Nov 9, 2014 at 9:30 PM | Registered CommenterMartin A

Michael hart

Feel free to go elsewhere. Your denial of the validity of the measurements I intend to use means that anything I say here will be irrelevant to you.

Nov 9, 2014 at 6:19 PM | Unregistered CommenterEntropic man

??? Most peculiar. Even by your standards.

When you deign to tell us about "the measurements I intend to use", then I'll consider reading them. I don't even see anything worth denying at the moment.

Nov 9, 2014 at 10:24 PM | Unregistered Commentermichael hart

Martin, it is not that I disagree about any particular number (27 was just an example), but that they seem to combine accuracy and precision into a percentage (of what?) in a way that is not useful. Some of those quantities are highly accurate and highly precise (emission spectra). Others might be quite inaccurate (like ocean levels) but reasonably precise (i.e. repeatable). And of course, even inaccurate/imprecise values can still be useful, as trends can still be determined.

Nov 9, 2014 at 10:36 PM | Unregistered CommenterRaff

And you start a "discussion" thread with my name, and a bald assertion about what I think. You then invite me to go elsewhere because what I think will make the thread irrelevant to me? Have you been drinking (excessively), EM?

Nov 9, 2014 at 10:49 PM | Unregistered Commentermichael hart

"As I said to michael hart, if you don't believe the measurements, this thread won't be relevant to you. I take the data and their confidence limits at face value and calculate accordingly. They tell a coherent story, which ,I would not expect from bad data."

If you find it difficult to address the points I've made to you, that's fine. But I've asked you two simple questions, one is what test would prove global warming - I don't need the measurements just what it is that would prove it for you.

On unthreaded you, quite rightly, said a hypothesis could be confirmed by its ability to forecast future events. Naturally, I'd assumed that the less than glorious performance of the climate models would have deterred you from quoting them, and you didn't. For my second question it's easy, you believe in a theory, 9000% of the scientists on the planet agree with you (I might have exaggerated), so please tell us what you and the 9000% of the scientists on the planet would regard as falsifying the theory. It's not that difficult, but I have to show some sympathy for your predicament because the real experts are waffling more than you are.

Nov 9, 2014 at 11:10 PM | Unregistered Commentergeronimo

Falsifying AGW is easy. If CO2 levels were to start falling despite BAU emissions, or if global average temperatures were to fall for a few decades, or glaciers were to start advancing instead of retreating - any of these would mean that the idea that human CO2 emissions would cause temperatures to rise was seriously wrong. You can probably think of (or google for) others.

Nov 10, 2014 at 12:10 AM | Unregistered CommenterRaff

geronimo

This thread started with the question of hypothesis testing in climatology. I have already made one simple test and demonstrated that one prediction from theory is not falsified by observation.

I'm interested here in continuing that theme rather than getting sidetracked by more general discussion.

Note that I took one specific prediction from theory and tested it against observation. I am not testing the whole thing at once. You asked what convinced me. It is the way my own calculations confirm that individual lines of evidence from climatology form a consistent whole. My future calculations here will illustrate this.

It would be much more useful to this discussion if you could critique my 8.33 post.

Nov 10, 2014 at 12:21 AM | Unregistered CommenterEntropic man

Raff

Falsifying AGW is easy. If CO2 levels were to start falling despite BAU emissions, or if global average temperatures were to fall for a few decades, or glaciers were to start advancing instead of retreating - any of these would mean that the idea that human CO2 emissions would cause temperatures to rise was seriously wrong


How about not rising for a couple of decades? You realise the implications of using OR rather than AND?

Nov 10, 2014 at 8:28 AM | Unregistered CommentersandyS

SandyS

The pause, if it ever existed, may be over. There is a flat spot n the GISS 5-year average. Every year since 2003 has a 5-year anomaly average of 0.57 0.58 or 0.59.

After 9 months the 2014 average is at 0.68 and rising. This would make the most recent 5-year average 0.61. Both are record values.

The topic of this thread is hypothesis testing. The "sceptic" hypothesis is that warming has stopped. AGW predict an ongoing long term rise underlying short term variation.

If the pause is ending, this would support AGW.

Nov 10, 2014 at 10:06 AM | Unregistered CommenterEntropic man

This would make the most recent 5-year average 0.61. Both are record values.

So 0.02C over the 0.59 C in 2003 to 2011 Anomaly with Base: 1951-1980 is an increase worth saying the pause is over !!!!!

EM you have surpassed yourself in crass comments with this one.

Nov 10, 2014 at 10:33 AM | Registered CommenterBreath of Fresh Air

Breath of fresh air

The 5 year average centred on 1998, the first of the"18 year pause" was 0.44; the 5- year average centred on 2012 will exceed 0.61.

I would say that a change from 0.44 to 0.61 would be sufficient to falsify the pause hypothesis.

Nov 10, 2014 at 11:35 AM | Unregistered CommenterEntropic man

Geronimo

Hans Custer suggests ten ways to falsify AGW here .

Nov 10, 2014 at 11:40 AM | Unregistered CommenterEntropic man

Something for which there is no evidence does not require falsifying.

Nov 10, 2014 at 12:06 PM | Registered CommenterMartin A

There is of course an easy way to test the AGW hypothesis - continue the so-called 'BAU' and look back after ~100-150 years. Even here, the IPCC has made up its range of projections to encompass such a wide range that even if natural variability were to result in an amount of change that is co-incidentally the same as predicted, the prediction would take credit.

The only surefire method of prediction success would be to see a temperature change that is both substantial and rapid - sort of like a true hockey stick - within the instrumental record.

Nov 10, 2014 at 12:09 PM | Registered Commentershub

The "sceptic" hypothesis is that warming has stopped. AGW predict an ongoing long term rise underlying short term variation.

If the pause is ending, this would support AGW.

Nov 10, 2014 at 10:06 AM | Unregistered CommenterEntropic man

You still don't get it. There is no single "sceptic's" hypothesis. Temperatures could easily resume warming at a gentle rate and the model predictions would still be wrong. Of course, if they had predicted only gentle warming at the out-set nobody would have been alarmed.

Nov 10, 2014 at 12:48 PM | Unregistered Commentermichael hart

Entropic man
Maybe, possibly, might be.

If ifs and ands were pots and pans there''d be no need for tinkers.

Maybe is not is over. Therefore it is just as likely to continue, or even become a decline. Therefore the questions to Raff are still valid, and do you think it is an OR or An AND?

Michael hart, exactly.

Nov 10, 2014 at 1:47 PM | Unregistered CommentersandyS