Maggie's Farm

Hi Zachariel,

An interesting set of posts, well and calmly expressed.

I think that a lot of doubt regarding the "climate change" debate is rightly focussed on the fact that data is highly inadequate to make accurate predictions. In the 1970s there was concern about "global cooling", which evolved to "global warming" and which is now labelled "climate change" just in case the direction of change out-foxes humanity again.

There are a lot of observable factors that might contribute to any changes and many measures are intuitively intrinsically hopeless to the scientifically-literate, such as "global average temperature" and the measurement of atmospheric trace gases that are an intrinsic part of our respiratory system. The process of science is only partially about getting accurate data. It is also necessary to perform very thorough analysis of how the many different observable factors act in isolation and in combination. In any natural data, there are many quasi-invariant long term effects that can be observed and I am unconvinced that the body of climate science is combining their understanding of specialist "components" into a full and correct model of the factors that influence the regional temperatures of the world that is fit for deciding policy options.

And this is where the whole thing dies. It becomes a political issue with major headlines about how "something must be done!!!" and supported by many people who are largely scientifically unaware and also by politicians who may have legal/economics/think-tank backgrounds but who cannot readily access the science independent of expert summaries.

On many issues, this is not a horrible problem. However, "green" influence in Germany has caused that country to plough many billions of Euros into subsidies for non-otherwise commercially viable alternatives and shut down nuclear. There is a very real danger of supply problems in the German energy grid in the near future as a result of this choice. (As I understand it, German nuclear has always been heavily subsidised too. But it is dependable and predictable and the renewables there are frankly not.)

It is the calls for extremely expensive action based on incomplete scientific understanding of a massively complex problem that worry most rationally sceptical people on this issue.

What if governments in the 1970s had spared no effort to halt the cooling of the planet based on mistaken science?

All the best,

Paddy

Dear Zachriel,

Thank you for your reply.

Where to start?

First-off, "probability ranges" are not the same as just showing the actual data that exists. If you look at estimates of "mean global temperature" they generally have some kind of trend smoothing presented, but do not show the variance around those means. At the surface it is -40 C in Moscow when it is +40 C in Sydney. Between these two points the mean is zero. At altitude, there is also a large dispersion. Even if you take evenly-distributed geographic locations and try to adjust for latitude and daylight hours and absolute numbers of real sample points: 1) there is a huge hole in the number of samples that are collected reliably at regular intervals. These are estimated through extrapolation. 2) The standard deviation is large around these data points. 3) The extremes of temperature are possibly more important than a Gaussian distribution of the statistics might allow for.

Try plotting the mean average temperature in one particular location over one day and add plus/minus bars to the extremities and the strength of movement in the mean is not so certain.

The graphs that are published are typically mean of means of means calculations: in each location there is a mean temperature every day, with error bars. Over the year you have a mean temperature per location over that year and you then have the combination of all locations. There are many ways of combining these data points into a number and the absolute and statistical variation over all points is very high, at least an order of magnitude higher than the range of values of the single mean presented. I call this measure "hopeless" because the demonstration of a rise in this data has huge variance around the derived mean values that dwarf the measurement. (The intuition is based on years of experience of statistical analysis on interesting scientific and engineering problems.) Using the same statistical probability techniques between 1950 and 1970 would point to a future fall in temperature based on the data alone.

It is a similar story for CO2 and other pollutant estimates. Even trying to place a figure on the levels of human contribution is hard to separate from natural events.

Yes of course your summary of the effects of pollutant classes makes sense, but the data measures are so fragile. It is imperative to collect much more and derive proper metrics in combination with all factors before we make any crazy decisions that we justify using science. The data used does not offer an acceptable confidence level to make drastic policy changes, but perhaps merits serious organised scientific investigation.

Regarding terminology, my main point is that dreadful decisions have been made based on popular perception of issues, based on what become propaganda terms used by people who "believe" in the scientific consensus. This is quite separate from terminology used in honest scientific publications. This phenomenon has been observed in other disciplines: "Artificial Intelligence" is now termed "Cognitive Systems Research", for example. The popular terminology and language used by lobbyists has certainly changed along the lines I have indicated as one piece of terminology becomes unfashionable. The terms are not reserved scientific terms at all.

We are commissioning research that "points" in a certain direction for future research, but offers us so far no reliable proof of causes and effects of the climate system. When I refer to the whole thing dying off, I refer to the science ultimately being pointless if it is misused. It is not read "as is", which is that we should investigate further. It is instead taken as done that these things are happening on the basis of ongoing incomplete research.

Terrible decisions have been made politically on the basis of incomplete research.

Dear Zachriel,

I thank you for your reply, but I fear that either I am miscommunicating the point, or you are not picking it up.

Let's ignore tedious point by point lists for now and let's focus on a particular paper on the link that you kindly provided from Berkeley:

http://www.scitechnol.com/GIGS/GIGS-1-103.php

This is rigorous scientific publication. Basically, what it says is: "well, we tried this out and this is the report". The authors explain all of the many mathematical assumptions and limitations very openly here, necessary because of their own admittedly incomplete understanding of the problem. This paper is only from March and explains how pauce an understanding of the underlying problem even experts/professionals have.

They are not at all offering real probability analysis on accuracy of real data. Rather, they are making assumptions about an intractable problem, having a go at modelling it and publishing the results of the model.

I suggest that you read the paper section on uncertainty analysis. They describe very clearly the difficulty in assessing this problem. The 95% "confidence" does not mean that it is 95% likely to be true, but rather that by hypothesis generation and testing according to their most extensive and well documented assumptions there is a 95% probability that the simplified data points according to the model lie within that range of expectation. This is a fundamentally different pair of claims that any scientist should find distinguishable.

A key assumption is that if they come across a value that is un"expected"in a statistical sense, they deliberately weight it down to improve the estimate confidence. This is allowed in a primitive model based on standard hypothesis testing, but it does not address the specific objections I already published for you above.

One last thing is the propagation of uncertainties between independent observations of different phenomena. If you have two processes that you are trying to model independently and if you do not have the luxury of making substantial assumptions without quantifying the impact of the assumptions on the result stability over a few thousand data points, the combination of these diverse effects can produce very low quality outcomes.

Near every academic paper has a positive outlook. However, an alternative reading of this work is that if the same group were to have non-essential assumptions disallowed, error/variance would be substantially higher. Similarly if we were to examine production of CO2 by human activity rigorously from first principles. Combining these effects outwith primitive assumption based models is clearly a long way from justifying major policy changes based on this science.

Best,

P

Oh! Here's a list!

Paddy: If you look at estimates of "mean global temperature" they generally have some kind of trend smoothing presented, but do not show the variance around those means.

Z: In scientific work, of course they do.

[b]Paddy:/[b] Not in the link you sent me! They reduce it to a confidence interval based on a model.

Paddy: 1) 2) 3)

Do you think the statisticians and scientists working in climate science are unaware of these issues? You might check out the Berkeley Earth Surface Temperature project, which recently reanalyzed the temperature record using new techniques.
http://berkeleyearth.org/results-summary/

[b]Paddy:/[b] No. Thank you for the link though, because it entirely backs up my previous point.

Paddy: (The intuition is based on years of experience of statistical analysis on interesting scientific and engineering problems.)

Then you might consider publishing specific objections, rather than your general comments here.

[b]Paddy:/[b] I just did...

Paddy: Using the same statistical probability techniques between 1950 and 1970 would point to a future fall in temperature based on the data alone.

Climate science isn't based on mere correlation, but is based on known mechanisms. The basic theory is a century old, though originally the theory wasn't based on anthropogenic forcing, but ancient climate change.

[b]Paddy:/[b] Well, yes in general, but the authors of the paper on Mean Global Temperature under discussion actually use correlation measures rather than covariance.

Paddy: Even trying to place a figure on the levels of human contribution is hard to separate from natural events.

That's actually not that difficult because the human contribution is very substantial and easily estimated. However, one interesting study uses carbon isotopes, which are different in fossil fuels than from other sources. See Ghosh & Brand, Stable isotope ratio mass spectrometry in global climate change research, International Journal of Mass Spectrometry 2003.

[b]Paddy:/[b] Thanks for the ref. Human CO2 surplus production is not easily distinguished from nature. It is a trace gas in the atmosphere and contains substantial regional variations.

Paddy: It is imperative to collect much more and derive proper metrics in combination with all factors ...

There are extensive efforts underway. The key policy question is called climate sensitivity, which will determine how quickly anthropogenic warming will occur, and how severe it will be. Most estimates are 2-5°C per doubling of CO2 with 95% certainty. The followup question is what this warming will do to regional climates, which is a much more difficult problem.

[b]Paddy:/[b] "Certainty" as defined in statistical hypothesis testing of simplistic models that can't even model one factor, let alone combine many factors correctly. The followup question is astute and key to the whole problem.

Paddy: ... before we make any crazy decisions that we justify using science.

The lag time is such, and the costs are such, that small actions now will have a greater effect than much larger actions later. In other words, it may make sense to move forward now, but certainly, you don't want to disrupt the economic system. It has to be a planned, gradual transition. Continued growth and development is essential to confronting the challenge of climate change while providing people with a better future.

[b]Paddy:/[b] A balanced comment, but it backs up my point exactly. There is certain scientific evidence out there"pointing" toward climate change. The reaction we need is to conduct proper research, not go mad and enter the domain of speculation plucked out of the air and waste money based on unscientific conclusions.

"If a meteorite were to strike the planet tomorrow 3 billion people might die! My computer model based on THE BEST ASSUMPTIONS WE CAN MODEL TODAY says that there is a 95% confidence interval that this may happen within 100 years. Accordingly the world should devote 5% of spending to an anti-meteor ring of laser-firing satellites. Gradually, before the meteor hits us."