Guest post by Steve Goddard
Archimedes had his eureka moment while sitting in the bathtub. Newton made a great discovery sitting under an apple tree. Szilárd discovered nuclear fission while sitting at a red light.
There was a time when observation was considered an important part of science. Climate science has gone the opposite direction, with key players rejecting observation when reality disagrees with computer models and statistics. Well known examples include making the MWP disappear, and claiming that temperatures continue to rise according to IPCC projections – in spite of all evidence to the contrary.
Here is a simple exercise to demonstrate how absurd this has become. Suppose you are in a geography class and are asked to measure the height of one of the hills in the Appalachian Plateau Cross Section below.
Image from Dr. Robert Whisonant, Department of Geology, Radford University
How would you go about doing it? You would visually identify the lowest point in the adjacent valley, the highest point on the hill, and subtract the difference. Dividing that by the horizontal distance between those two points would give you the average slope. However, some in the climate science community would argue that is “cherry picking” the data.
They might argue that the average slope across the plateau is zero, therefore there are no hills.
Or they might argue that the average slope across the entire graph is negative, so the cross section represents only a downwards slope. Both interpretations are ridiculous. One could just as easily say that there are no mountains on earth, because the average slope of the earth’s surface is flat.
Now lets apply the same logic to the graph of Northern Hemisphere snow cover.
It is abundantly clear that there are “peaks” on the left and right side of the graph, and that there is a “valley” in the middle. It is abundantly clear that there is a “hill” from 1989-2010. Can we infer that snow cover will continue to increase? Of course not. But it is ridiculous to claim that snow extent has not risen since 1989, based on the logic that the linear trend from 1967-2010 is neutral. It is an abuse of statistics, defies the scientific method, and is a perversion of what science is supposed to be.
Tamino objects to the graph below because it has “less than 90% confidence” using his self-concocted “cherry picking” analysis.
So what is wrong with his analysis? Firstly, 85% would be a pretty good number for betting. A good gambler would bet on 55%. Secondly, the confidence number is used for predicting future trends. There is 100% confidence that the trend from 1989-2010 is upwards. He is simply attempting to obfuscate the obvious fact that the climate models were wrong.
Science is for everyone, not just the elite who collect government grant money. I’m tired of my children’s science education being controlled by people with a political agenda.
AlexB (21:48:12) :
Robert (21:25:27) :
vigilantfish (16:23:54) :
AlexB (17:35:46) :
AlexB (21:48:12) :
Re: Leif Svalgaard (20:30:22)
‘Scientific hypotheses can be elevated to theories by fact, and theories can have differing degrees of empirical content but a scientific theory can never ever be a fact.’
Steve Goddard (21:54:18) :
You have the scientific method round the back of your neck, as Nottingham people say it when someone tells something in reverse.
A theory is generated by becoming aware through study of a gap or an anomaly or an interesting possibility in the research in a particular field. This theory then generates hypotheses (with sufficient ingenuity) that can be tested, supported or falsified. Experiments or observations are carried out and statistical tests used to find out if the results found are likely to be by chance. If the conventional p-value is achieved then it is likely that the hypotheses are reliably supported as not due to chance-if not, the null hypothesis or hypotheses has to be accepted and you find that your theory does not stand the test of rigorous examination by the scientific method. Time to start again with another bright idea.
The word ‘fact’ should never be part of a scientist’s vocabulary because nothing is an established ‘fact’-repeat, nothing! I point you in the direction of the way this word is used in the TV program ‘The Office’-do any of you wish to be associated with that? All information used in science is temporary and contingent, likely to be overturned at any moment by fresh information that has gone through the testing process. The only people who cling to outdated scientific findings or an overall paradigm are crusty old professors who have an emotional investment in them.
‘Simple explanation from William M Briggs, statistician.
Let’s ignore statistics and turn to plain English.
Suppose, fifteen years ago the temperature (of whatever kind of series you like: global mean, Topeka airport maximums, etc.) was 10 C. And now it is 11 C. Has warming occurred? Yes! There is no other answer. It has increased.’
The point is ‘Is the increase statistically significant?’ A simple numerical increase means nothing, it can be by chance and could just as easily reverse itself. As well, the amount of the observed increase has to be enough to be statistically significant. I drummed into my students that you cannot support or falsify an hypothesis with descriptive statistics; that could only be done with inferential tests.
All the statistical tests referred to in these comments are based on correlation; even regression is ultimately based on correlation and however you dress it up the old saying ‘Correlation is not causation’ holds true. There was a mention of Chi-squared but that is one of the weakest statistical tests. Climate science seems to be difficult to pin down and torture with the stronger inferential tests, on the results of which some reliance could be placed. Correlatory results provide some very thin ice on which I would not like to stand.
The latter part of my comment, about ‘expanding bags’ and alarmists having it both ways every which way, seems to have been ignored but it is at the core of my complaint about them: no testable hypotheses can be generated from a theory that accepts all information as supporting its claims. Let them have the courage to design a set of hypotheses that may be tested by strong inferential statistical tests and thereby stand or fall. If those hypotheses fall then they must close their case, slink away into the sunset and the rest will be blessed silence.
Steve,
Your anaolgy is useful, but probably incomplete.
The task before us is to determine if the section of the range of hills that we have sampled to date (starting from the West) is actually overlayed on some underlying, larger or longer, hill or valley. I think we can all safely assume that this range of hills cannot continue to rise (or fall) into infinity.
Robert (17:38:37) :
Shot yourself, not in the foot but in the head.
You obviously missed the “TEMPERATURE LEADING BY 1000 YEARS”.
The key to good science is actually knowing what the most important things to measure ARE.
Experimental science usually starts with measuring/monitoring things where instruments are already around. You know the sort of thing. Looking at Arctic Ice extent for a year or so – your schoolkids could do that as a project, couldn’t they?
Next is to try and discern trends, which again is taught using long-standing records. You could look at CET, couldn’t you? You could look at SOI?? You learn some maths, some graph plotting and you hope that some start to sniff out some features.
Then, and only then, do you come to prediction. You usually assume that the past is a guide to the future, because if not, how do you predict with accuracy? So you start trying to predict next years’ winter snowfall, next winters’ ice maximum etc etc. Some will be right, some will be wrong. You eliminate a bunch of theories in Year 1 and a few stand up.
Now then: there will be some who, by luck, chance, design or intuition think they might be rumbling nature. They make predictions for 7 – 10 years and they’re on the money. Who IS this genius?? What do they know that I don’t?? And, once again, they are admired because they called it right for a while.
Then something happens to bust ’em out. Maybe the PDO shifted? Maybe AMO changed? AO, NAO, IPWP?? Maybe the data sets weren’t consistent or accurate? Hadn’t factored that in, had they?? Bummer. Hero to zero in one failed crop season. The dangers of hype and infallibility…
In each iteration, a subset of key parameters are flushed out which contribute to understanding and, hence, prediction skill. In each iteration, new measurement tools are created, false theories are put forward, tested and fail. In some rounds, salesmen are lauded before the theory they are selling has been substantiated. Religion has entered the fray…..
And so a crisis of faith comes along sooner or later, where two polar opposites collide. And out of that mess can either come a war and a Dark Age or yet deeper understanding of systems so subtle, so complex yet showing, to the discerning, clear signposts to reach the next plateau of understanding.
It took 10 minutes to write this, but 25 years of a journey to experience the components of it.
That’s why you don’t learn to be a scientist reading a blog or a paper. Because they distil 10 years of struggle into two columns of print.
‘I am a father’ encapsulates meeting, romancing, courting, marrying and starting a family. Four words. 10 years of honing who you are and what you became.
It’s the same in science.
And it’s a crying shame that no science textbooks that I was deputed to read explained science quite like that. And no school education programme nurtured scientific enquiry in that way.
That’s life.
Maybe the 21st century will prepare the world to embrace the reality of long-term science in the 22nd, eh?
Philosophers have spent their lives and written millions of pages about “facts” for a reason .
There is no straightforward definition accepted by everybody about what a “fact” is .
Actually it covers a very large spectrum going from what everybody can detect by his senses without needing any interpretation frame to completely belief driven convictions only based on interpretation frames .
If you look at the sun with naked eye you will become blind is a fact that everybody agrees with , can validate (if he dares) and doesn’t need any interpretation .
.
Scientific “facts” are sometimes tricky because they critically depend on a theory to be valid . If the theory is valid , its mathematical being allows inferences and predictions . If the thing predicted is observed , it increases our belief that the theory is correct and tends to upgrade its statements to “facts” .
However one has to be clear that this kind of “facts” are only conditional and the history of science is full of such “facts” that stopped being facts once the theory was proven to be wrong .
A good example are the black holes .
Nobody had flown around one , thrown objects in it , measured its temperature and otherwise experimented LOCALLY with them .
Yet it is a not so difficult to derive their existence and properties from the GR .
So depending on the faith you have in the validity of the GR equations , you will say that the black holes are a fact or not .
If you happen to interpret some radiation observations in terms of black holes and it sticks , your faith will increase even if you have still not experimented locally with one .
Perhaps the GR will be busted one day . And that day the black holes can stop being the “fact” they are today .
.
AGW is still far even from that hypothetical “fact” stage because there is not a mathematically formulated theory that allows inferences and predictions . Sure computer simulations make a kind of probabilistic scenarios but that is not a fully formulated predictive theory .
And no , absorption/emission properties of CO2 or vague correlations are not enough to draw even qualitative conclusions .
Sofar AGW is just one size fits all working hypothesis and no fact .
It would be interesting to do a little Hurst analysis on any of the data. It would basically show the length of time a model is good for (if it is good for anything) before it goes to pot. Weather like the Nile floods (which were a product of weather) are a mathematically chaotic process.
The Golden Calf of many a fake science, as the so called climate science, is about to fall down…The “turn of the screw” is here!
Atticus,
Learn how to read a graph before commenting.
http://wattsupwiththat.files.wordpress.com/2010/02/dec-feb_snow_ext.png?w=510&h=291&h=291
Winter snow cover has increased over the last 20 years. Anyone with a high school science or math education can see that.
Comments from the director of the Rutgers Snow Lab
http://www.google.com/hostednews/ap/article/ALeqM5g1jo1gT0843vxrD4oRUd1Ufm4F5AD9DRBO880
Simple explanation from William M Briggs, statistician.
http://wmbriggs.com/blog/?p=1958
Richard,
Given that the earth is billions of years old and is neither covered with snow, nor snow free – it is safe to assume that the long term trend of snow cover is flat and has a slope of zero. That tells us nothing about the last 20 years.
Anthony Fallone,
At no point have I ever made any attempt to predict the future of snow cover. I have not made any hypothesis about a cause and effect relationship between the year and the snow cover. I have simply observed the undeniable observation that winter snow cover has increased over the last 20 years. Statistics has nothing to do with it.
Had I made a prediction of how further increases in the X-Axis affect the Y-Axis, then your argument might be valid. But that is not the case.
h2o273kk9 (22:37:24) :
does Huygen’s theory mean that light is in “fact” a wave or just that it acts like you would expect a wave to act.
If it looks like a duck, quacks like a duck, etc.
At some point we attach descriptive words to phenomena. Anything that has a ‘wave length’ is called a wave.. ‘is in fact’ is a different question, namely what is reality? The ‘fact’ I was referring to is that light has a well-defined wave length.
Michael Larkin (01:32:46) :
IMO, evolution over long periods of time is an observable fact: just look at the fossil record. However, the absence of gradualism in that, at least to my mind, challenges neo-Darwinism as providing a complete theory.
Things can be facts without being fully described by a ‘complete’ theory.
Dr Anthony Fallone (02:25:00) :
‘Scientific hypotheses can be elevated to theories by fact, and theories can have differing degrees of empirical content but a scientific theory can never ever be a fact.’
Nobody said that or, at least, implied that. The theory is about facts or tries to describe facts. Kepler’s laws are not facts, but a description of observed positions of the planets [those were the facts], and eventually General Relativity is a theory of and explanation for the fact of gravity. Darwin’s work was mainly a description of the fact of evolution with a theory attached that gave a plausible explanation for that fact.
That the Earth is round is a fact, which can be explained by Newton’s laws [which also explain why the Earth is not a sphere, but has a flattening brought about by the rotation of the Earth].
Gravity is a fact with or without Newton. Evolution is a fact with or without Darwin. Both men provided a description and a theory of the underlying facts.
Nice, simple, straight forward insight.
I’d have thought that any measurement is in the end about whether it leads to something testable and practical. “There are no hills” is obviously an impractical finding.
The focus on the last 200 years or so of surface temps, which ignores the recent 10 or 15 years as “too short”, and ignores anything beyond 300 as “too long ago to matter nor understand” — that focus on the last 200 years should provide us with a hypothesis that we can test and use for practical problem solving. All it seems to have done is produce models forecasting scenarios 100 years out that we can’t test. If we can’t test then what’s the point?
If we can’t test, it is not science. So sure, “we have to do something” — I’d just as well rely on my own intuition. That’s what most people seem to do. You can’t publish intuition, but you can act on it. My intuition says there’s too many good things that might happen in the future which we don’t want to jeopardise by imposing silly resource cuts.
h2o273kk9 (23:31:32) :
And this 30 year time frame was established where, precisely? I must have missed that meeting.
You were not even born then. Long before there was a climate debate, meteorologists [formalized by The World Meteorological Organization (WMO)] established and generally agreed upon the dividing line between weather and climate being at 30 years. This is, of course, arbitrary in a sense. Perhaps it should be 31 years or 29, but the order of magnitude is about right, or so they found empirically, and have agreed upon as a useful standard or norm in order to be able to compare data from different providers.
Very late coming to this party 🙂
Robert wrote:
“A confidence interval is about distinguishing a random distribution from a pattern. By convention, you need to be 95% confident in your trend in order to reject the null hypothesis. 90% is on the bleeding edge of acceptable. Less than 90% is not statistically significant by any measure.”
Robert, you need to attend one of my Intro Physics course lab sessions where students routinely show me how they have fit a straight line to their data and the computer spits out a r^2 value of 95%. They then confidently tell me how that has to be the correct result because of that r^2 value. They, like you, ignore what their eyes are telling them about the real data trend which can be a power function, etc. They quickly learn that the 95% confidence interval is a deceptive statistic and being dependent on it leads one to conclusions that are not valid.
Steve,
The discussion is interesting and it does show us something about how all things climate can be looked at.
Obviously, snow extend has risen. The data shows that there’s more snow. But is it noise or a trend? If it were a stock chart that I was looking at, I’d say, “More noise than trend” and move on to a clearer chart.
The funny thing is, that’s what climate and weather really are, a whole lot of noise and not much trend. It both validates Tamino’s argument and supports the notion of AGW Alarmism as utter foolishness over normal and natural variation. MOST of the climate data I see charted looks like a stock I would not want to trade. There’s no “robust” trend, especially when you adjust for normal climate cycles (ENSO, etc.).
Mark
Ref – R. Gates (16:29:54) :
“..on average we are seeing more moisture and thus more heat and evaporation from the oceans during these months. How that discredits AGW in any way is beyond me…”
________________________
Seems reasonable, until you beg the question “How that discredits AGW in any way is beyond me..”. AGW doesn’t forecast another 80K year glacial phase, just the opposite, so its likely not AGW “if” the world’s ice increases.
_______________________________________
Observation to all –
Academics is a fine and worthwhile endeavour. To read through all the bits and pieces of discourse here at WUWT is a fascinating way for me to waste time while waiting for the Grim Reaper, but the question of “Are We or Aren’t We causing changes in the Earth’s climate?” is just that “academic”. Just as fascinating –often more so– as the question about the accuracy of the comments, is the so very human emotional component of the discourse. Plato might not hang around here too long, but everyone of you adds a special something to the discussion in my book. Thank you (and I do mean everyone). You make my day:-)
Every sufficiently long noisy signal will contain runs that look like a signal but are in fact noise. Nobody disagrees that Winter snow cover has, on average, increased since 1989. It has and that is a fact, and as you correctly point out, does not constitute a prediction that the increase will continue. I think the disagreement is whether this increase represents noise or signal. There are objective tests for the hypothesis that a rising or falling signal is embedded in a subset of the sequence. Anyone can loook them up.
In the context of this discussion, when you say, “Winter snow cover has increased over the last 20 years,” I hear an assertion that the time sequence of N. Hemisphere Winter snow cover contains a rising signal during the last 20 years that is distinct from the superimposed noise. Otherwise, why say it? It would serve no purpose to say that this or that subset of pure noise is rising or falling.
Either you are making an observation that the last 20 years of noise has randomly produced a rising trend, or you are saying that there is a rising signal embedded in the last 20 years of noisy data.
Which is it?
If it is the latter, what recognized objective test have you applied to validate your observation. If it is the former, why even mention it?
And just to be clear, we are both playing for the same team in the AGW debate. I am just asking you to step up your game so that our team can win sooner and more decisively.
Caveman,
The only reason to establish statistical significance would be to establish a cause and effect relationship between the X-axis (time) and the Y-axis (snow extent). That would be an important piece of information if we were going to try to predict ice extent moving forwards. I am making no attempt to do that however.
Regardless, Tamino calculated 99% confidence before he applied his undocumented “cherry picking” test.
Speaking of cherry picking, a favorite study at Copenhagen was this Greenland melt study, which showed that Greenland melted a lot between 2003 and 2007.
http://www.spiegel.de/international/world/0,1518,661192,00.html
I’m really surprised that Tamino did not apply his cherry picking test to that.
Steve Goddard (07:53:35) :
The only reason to establish statistical significance would be to establish a cause and effect relationship between the X-axis (time) and the Y-axis (snow extent).
plot the snow cover data on the SAME graph as the model graphs you have been showing, and plot ALL the data for both.
Back in the early 90’s when I was a fresh faced Navy Wx observer,we took old school Airway observations. Then we went to Metar obs, a flawed and not as detailed observation method. Now, except in rare cases, all obs are ASOS generated which won’t report cloud decks below 10K.
I miss the old obs. The taking of them is a lost art and much valuable info is lost.expeciallly cloud type.
JLKrueger (20:41:48) :
davidmhoffer (17:49:17) :
That’s how computers see data and that’s why it is so easy to take big chunks of data and get answers that are technicaly accurate and totaly meaningless. >
BINGO! A sure nomination for quote of the week! It was one of the first things drilled into us when I was learning to be an ORSA (Operations Research Systems Analyst). It’s also the first lesson forgotten as we become enamored with our fancy toolkits.
Thanks JLK. I’ve been selling technology for 30 years, much of it to heavy weight researchers. I keep using the Autocad/Stairwell story to show that when we rely on computers, we become limited by the capabilities of the tool we are using.
Along the way I had some amusing stories. The average researcher is far better equipped to deal with computers than they were 20+ years ago, but here’s a few fun ones:
Them; This 24 cpu computer you sold us runs exactly as fast as the 1 cpu computer we had before
Me; OK, remember that conversation we had about converting your code from single threaded to multi-threaded? Did you do that?
Them; This RISC computer you sold us corrupts the data every time we move it from our Intel computer.
Me; OK, remember that conversation we had about byte ordering and that Intel and RISC use opposite byte ordering when reading and writing data? Did you account for that?
But the funniest one (I could NOT make this up)
Them; We bought a blade server farm with 4 cpu blades from vendor X and it ran 1/4 as fast as it should. We bought a blade server farm with 8 cpu blades from vendor Y, and its even worse, it only runs 1/8 as fast as it should. We’d like to try a blade server farm from your company.
Me; Uhm… I’ve reviewed the list of apps you are running, they are all single threaded, so they can only run on 1 cpu of each blade. My product won’t act any different. What I suggest is that you run the FREE version of VMWare on your blades, virtualize each blade into one O/S instance per cpu, and then run your application O/S on top of that. No cost to you and you can use the h/w you already have. Or you can spend $2 million with me for a product that will have the same problem as the other two.
Them; (I did NOT make this up) I smell some sort of agenda here, I just can’t figure out what it is….
I’ve tried arguing this very same argument with AGW believers for some time now, but to no avail. Invariably, when you show that 1998-2009 shows no appreciable warming they will fall back on their “timescale insignificant” trope.
The point is, however, that length of time scale is less significant if you are dealing with cyclical changes. My example is a sine wave. if you have a sine wave with a wave length of 10cm you will get vastly different predictions of future trends with a sampling of anything less than 10cm. Furthermore, predicting the next 5cm of that wave really depends less on the length (below 10cm) and more on the where the sample is taken.
In a blind evaluation a 2cm sample at the apex of the curve would show a down trend while the 5cm sample on the up slope would incorrectly diagnose a rising trend for the next 5 cm. In this example the shorter trend is a better descriptor.
Leif,
Here is the image of the entire Rutgers winter data set (from this article.)
http://wattsupwiththat.files.wordpress.com/2010/02/nh_snow_extent_1967-2010.png?w=510&h=535
The overall trend is nearly flat since 1967, and we are currently at a near record maximum. Tell me, does that agree with model predictions of declining winter snow cover?
Robert, I read your first comment and I skipped to the end here. I didn’t put on my snowboarding gear, though.
If we accept only 90% confidence interval as *bleeding* edge
Then review the AR4 based on data of similar noise, the merger of multiple types of data and data sets with variable sampling rates and intervals and data populations,
to create first order LINEAR trends
for variable and possibly CYCLIC systems,
and call these first order linear trends climate models,
that they can predict reasonably to 50% into the future of the time that most of our data reasonably goes into the past,
to confidently identify that greenhouse gasses are the largest factor (such that we can discount others to simplify our model,)
and then have people go and attack OTHER people that suggest that 95% confidence interval would be necessary for refuting data,
I am 100% confident that I just crapped my pants.
It is like a Monty Python skit.
The Met Office used their climate models ten years ago to predict that “Snowfalls are now just a thing of the past”
http://www.independent.co.uk/environment/snowfalls-are-now-just-a-thing-of-the-past-724017.html
Now that the UK has had two consecutive very cold, snowy winters, the Met Office can declare that it “got better.”
[youtube=http://www.youtube.com/watch?v=fr8DIg3oHFI&hl=en_US&fs=1&]