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.
Steve Goddard (11:34:18) :
It isn’t realistic to do comparisons vs. the Precambrian.
Of course, but that is not the point. I was just discussing what your own Figure was purporting to show. Perhaps it was not the best one to bring up?
Now, how about that overplot [6th time] of Rudgers on Models?
Robert (11:34:56) :
95% confidence in rejecting the null hypothesis and an r^2 value are two different things. You need an n value to relate the one to the other.
n was known. r^2 was not, hence asking for r^2 is proper.
The claim was about how the observations and Models compare, so an appropriate graphs would have been a plot showing both, and a scatter plot showing observations versus models. For that last one, r^2 [and n] would be very illuminating. What I’m trying to convey is that there is a better [and time-honored, standard] way of comparing prediction and observations. Beats me why Steve is so averse to do that.
Leif,
The average from 1989-2000 was 44,500,000. The average from 2001-2010 is 45,600,000 . Seven of the top ten were after 2000. Eight of the bottom ten were prior to 2001.
No matter how you look at it, winter snow extent has increased since 1989, and is near a current record now. All your protesting isn’t going to change a thing.
Leif,
I don’t have the raw Columbia data to do a comparison against – do you?
But the point of this article is that such analysis is unnecessary at a qualitative level. Winter snow extent has been increasing while the models predicted it should be decreasing.
Steve Goddard (11:48:24) :
All your protesting isn’t going to change a thing.
I’m not protesting. I’m trying to make you present the data in a better way.
Steve: Using Excel or some other software to draw a linear fit or other trend line through some data doesn’t mean that the observed slope is meaningful. Some points will be above the line and some will be below. No problem, you say; that’s just noise in the data – and you’d be right. However, if these deviations are really noise, then you need to accept the fact that, BY CHANCE, some of the time, more of the upward deviations will be on the right side of the graph and more of the downward deviations will be on the left side of the graph. This will increase the observed slope of the graph. (You may not see this pattern in this data, but if it were possible to repeat the experiment many times, random noise is certainly going to show this pattern some of the time. Unfortunately, you won’t notice the unusual arrangement of deviations because the linear fit will be steeper and appear to be correct.) Based on the magnitude in the scatter of the data above and below the line and the number of data points, one can calculate a confidence interval for slope of your line. A confidence interval provides a context for interpreting the slope of your line. It is very tiresome seeing blogs jump to conclusions about linear fits without a confidence interval.
If you are using EXCEL, turn on the data analysis tools and select regression from the package. Looking at your 1989-2010 snow data, I find that the slope for your line is +97,000 mi^2/yr with a 95% confidence interval of 21,000-172,000 mi^2/yr. This is an 0.21% increase per year (or 21% per century if you translate into the usual timeframe for climate change). The 95% confidence interval is 0.05%-0.38% increase per year. So you have some confidence that there has been an increase in snow cover, but a very little idea (with a 7-fold confidence interval) of how fast or slow the increase is.
EXCEL lets you choose the confidence interval. Most scientists and scientific journals aren’t interested in placing bets on an 85% chance and having the conclusions reached by one out of every six papers be due to chance arrangement of noise in the data. So they traditionally look for >=95% confidence before allowing a conclusion in the abstract and are happier with >99%. [The IPCC’s “more likely than not”, “likely”, and “very likely” are political BS, not proper science.]
You should also look at the other output from the regression analysis. The plot of residuals shows no significant trends or correlation from one residual to another, showing that a linear fit is reasonable. The normal probability plot shows that much larger deviations from the mean are found in the few years when the snow cover is highest than when it is lowest. [A quick look at Descriptive Statistics shows that both skewness and kurtosis are significant, about than 3-fold the standard error estimated by SQRT(6/N) and SQRT(24/N) respectively]. This suggests that the data may not be normally distributed, and increasing skepticism that the 95% confidence interval derived on the assumption of a normal distribution of noise is wide enough.
You could repeat this analysis for the 22 years starting in 1967, 1968, 1969, etc.; giving you 23 chances to find a 22-year time frame where the snow coverage is increasing. Using a 95% confidence interval, you will probably find BY CHANCE an average of one 22-year period where the slope is significantly positive OR negative with 95% confidence. So you leave yourself open to a charge of cherry-picking data if you don’t have a particularly good reason for choosing 1989-2010. Let’s say you are interested in recent trends in snow cover – that provides a rational for ending in 2010 and not going all the way back to 1967. You still need a rational for picking 1989 as the starting year. So a scrupulous analysis might look at the trend for the last 15 and 25 years or the last 10, 20 and 30 years, hoping to demonstrate that the strength of your conclusion is independent of your choice of a starting date. Unfortunately, there IS something unusual in the finish date: You are doing this analysis because 2010 is an outlier with unusually heavy snowfall. So a scrupulous analysis would omit 2010 (which isn’t finished in any case). For the 20 years (picking a multiple of five without looking at the data) from 1990-2009, the change in snow cover is +0.10% per year (-0.05% – +0.26%; 95% confidence). (The t-stat is 1.38 which gives a 9% probability that the slope could be zero or less by chance.)
Is there a recent upward trend in snow coverage? Should you claim to have “proven” that the IPCC’s projections were wrong? With some statistical context, what do you feel like concluding?
A true “climate skeptic” should be as skeptical of data that agrees with one’s expectations as one is of the “scientific consensus”.
Re: Mike D. (Feb 21 13:46),
Get a real job, one the free market values and is willing to pay you for. Do “science” in your garage in your spare time. I’m tired of footing the bill and getting BS in return. Will society collapse if taxpayers stop funding “science”? I doubt it sincerely.
This caught my attention.
Depends what you mean by collapse. If there is no public funding for science and we go back to the philanthropically financed universities and only elite science, the most probable outcome is that we will find ourselves in a 19th century society. That would stop CO2 footprints for sure.
The great and accelerated progress in the technological world we live in has come from the funded by the public education and research, because a lot more brilliant people could have access to education. The same is true of the concurrent economic progress, the existence of a middle class and an eight hour day is tightly tied with this progress in technology.
Ideally, if one makes a science fiction extrapolation of technological advances, all humans will end as people of income lived in the 19th century, but served by robots, and the problems will be the problems of a leisure class : how to occupy oneself.
Though there is a lot in what you say about cliques and free loading if possible, one should not throw the baby out with the bathwater. The house should be put in order but that does not mean to stop public funding of universities and science.
Steve Goddard (11:34:18) :
Climate sensitivity needs to be calculated during more recent periods when oceans and life forms were more mature, like starting in the Cambrian Era.
A time-honored method is to try to investigate things in isolation [see e.g. http://www.leif.org/research/suipr699.pdf on how to separate and discover the contributions of the many different factors that together cause geomagnetic activity], if possible. It seems reasonable that by stripping away complicating factors that the effect of CO2 itself may show up.
“No matter how you look at it, winter snow extent has increased since 1989, and is near a current record now. All your protesting isn’t going to change a thing.”
At least a half-a-dozen people, not just Leif, have explained to you that there is no statistically significant trend. Has it occurred to you that perhaps there is a point here that you are missing?
Leif:
Thanks for the reference, which vindicates my observation that the idea of climate being defined by 30-year intervals must be of recent origin (remember, I’m a historian – 1960 is recent). As your first source states:
” The concept of a normal climate goes back to the first part of the 20th century. At that time, lasting to about 1960, it was generally believed that for all practical purposes climate could be considered constant, no matter how obvious year-to-year fluctuations might have been. On this basis meteorologist then decided to operate with an average or normal climate, defined by a 30-year period, called the normal period. Later, people became aware of the fact that climate is not constant, but undergo variations in itself.”
Although 1960 is mentioned as the last year in which people still held onto some notion of a constant climate, your source is still somewhat ambiguous as to exactly when the 30-year intervals were decided upon to measure changes in climate once the new paradigm was accepted.
vigilantfish (10:15:50) :
I would like to see a historical reference on this one.
I found Guttman’s article:
http://ams.allenpress.com/archive/1520-0477/70/6/pdf/i1520-0477-70-6-602.pdf
This is as authoritative as it gets.
Tamino’s back with some more charts, hoping to inundate with volume.
He attacked Lucy Skywalker for not showing the anomaly(correctly in my view), but then he starts off by showing total snow cover.
Then look at the next post, where he labels a sphere 2 dimensional and a circle one dimensonal.
“The n-sphere is the n-dimensional surface of an n+1-dimensional ball. The 1-sphere is just a circle; the 2-sphere is the ordinary sphere we’re all familiar with; “
vigilantfish (12:26:36) :
Thanks for the reference, which vindicates my observation that the idea of climate being defined by 30-year intervals must be of recent origin (remember, I’m a historian – 1960 is recent).
The concept goes back to 1872. Is that also recent? My point is that the 30-year interval is unrelated to the current debate.
http://wmbriggs.com/blog/?p=1958
Tamino should demonstrate that the climate model predictions of declining winter (January) snow cover are correct.
http://climate.rutgers.edu/snowcover/png/monthlyanom/nhland01.png
Frank,
Why do you call 2010 an outlier? 2008 had the third highest monthly extent on record and possibly higher than any in 2010. Should we throw out 2008 as well?
vigilantfish (12:26:36) :
the idea of climate being defined by 30-year intervals
If you read carefully you’ll see that the 30-year mean should be updated every 10 years. So, every 10 years we can have a new climate ‘assessment’, if you like.
Steve,
Clarification: African or European?
Mark
(bravely running away)
Dallas could set their all-time record for the snowiest winter ever if 1.9 inches falls in the city. That record was last set in the winter of 1977-78.
http://www.accuweather.com/regional-news-story.asp?partner=forecastfox®ion=southwestusnews
Mark Young,
You have to know those things when you are king.
@Robert (15:42:53) :
“Dave, I don’t know if your job at the Seven-Eleven doesn’t pay you enough, or what, but you seem obsessed with my phone and the fact that I have a climate science app on there. In fact, I look things up on all kinds of sources, all the time, if that helps you. We can’t all just repeat what a talking dog tells us.”
Thanks Robert. I sort of knew you’d go there and your comment about my job is about as accurate as I would expect you to be.
For reasons that you will never know, you have just made my life significantly easier and I thank you for that.
Anyway, that’s enough from me. Have to get back to work. Them there shelves ain’t going to stack themselves, you know – well, that’s what my dog is telling me…
Dave
Leif,
You can’t strip away oceans and vegetation from CO2 analysis, because they are responsible for the vast majority of CO2 emissions and absorption. Not to mention that the ocean holds almost all of the latent heat in the climate system.
Also, deserts get hot during the day and cold at night for a reason – the lack of vegetation. It wouldn’t make any sense to do a CO2 sensitivity analysis at a time before vegetation covered the planet.
Suppose you were in a geography class and you were asked the height of the hills?
That is like being asked how snowy the climate is.
Suppose you were in a geography class and asked whether the hills get higher towards the right?
That would be like asking whether the climate is changing.
And if you are trying to answer the second question, measuring the slope up one hill would be dishonest.
One has to be very careful when applying elementary statistics. The simplicity and convenience of elementary statistical tests is often attained at the price of some very stringent requirements for validity, and the “p-value” will be in error, often grossly so, when those requirements are violated.
For example, when determining the statistical significance of the slope of a line, you are not allowed to select a portion of the data for analysis based upon your visual inspection of the data. You can use all of the data, or you can randomly pick a starting point without looking at the data, but you cannot choose a starting point for the analysis based upon when you think the trend starts. If you violate this rule, the p-value obtained will be invalid, and likely smaller than the true value.
Now this doesn’t mean that you can’t statistically analyze data in which the slope actually changes at some point–it just means that you can’t apply the basic linear regression analysis that you find in all of the elementary texts–you have to apply a more sophisticated analysis that accounts properly for the additional degrees of freedom that this more complex statistical model introduces.