Studying Temperature Data Using the Language of Science

Hoosier wrote: “I would think that all of those 51 new excuses for the hidden warming should be supported with some proof. ... After all, the completely infallible computer model said so.”

Earlier in the discussion I believe you said there was “no evidence” of additional warming in the oceans. If you really wanted to know whether or not that was true, why didn’t you do your research? The fact that you keep asking is proof of it’s significance. But truth be told, you didn’t bother because you don’t want to know! Even though I made a point of looking for conclusions derived from direct measurements, this too, will all be rejected. This first one pertains only to the month of June of this past year:

Was June an anomaly or further evidence of a trend? The article below, citing actual data, contends that 2014 was the warmest on record. Note that above, the 2013 report pertaining to warming oceans is also based on actual data. Now, that study was limited to the Pacific & Indian oceans.

These last two make correlations between the warming of the oceans and other ramifications. This first one links the warming to sea-level rise:

So, direct measurements that are then used to establish a global mean. I will admit, if I’m reading that correctly, surprised to learn that the mere warming of the oceans accounts for 32% of the avg mean rise. This last one links the warming to the oceans ability to absorb CO2. Predicting that if this trend continues, and this has been said elsewhere by others, the oceans won’t be able to help us in future, right when we’ll need them the most. Again, this study was only at a given depth, on a small scale. It can be argued that this is a tentative finding. It is suggested that more studies will follow.

 

Thanks, and I'm sorry I missed that when you posted it. But it certainly deserves examination. Here's the relevant graph:



The basic problem here is that you're a geologist, and in geology the data don't move. When you look a a rock today, it looks just the same way it looked yesterday. And when there's no movement in the real world, any and all movement in the data must be due to error.

But unlike rocks, temperatures do actually change in the real world. That means that when you look at temperature data, some of that change is really due to real changes in temperature, and not due to error. And the problem with Mr. Science's graph above is that he has assumed that all, 100%, of the change in temperatures is due to error, and none, 0%, is due to actual changes in the actual temperature.

Oops.

Beyond that, Mr. Science is assuming the result he wishes to demonstrate (that being, temperature is not changing), which is circular reasoning.

Oops #2.

And even beyond that, in deriving his erroneous "errors", Mr. Science failed to account for trend, and failed to account for autocorrelation. Not only are these Oops #3 and Oops #4, it also betrays the fact that Mr. Science has no experience in time-series analysis. Of course that's not a sin, but when someone with no experience in the relevant branch of statistics presumes to point out "errors" made by those who spend their whole life in the field, all I can say is that Mr. Science is already in way, way over his head, and he's only making himself look stupid.

Meanwhile, we're still waiting on any real invalidation of a single datapoint, or of a single error range, in modern climatology.

O idiocy, thy name is Denierstan!

 

I'm a petroleum engineer, and my data moves every day. Every minute of every day. Sometimes it goes to zero. Sometimes open choke flow is constrained by tubulars. Folks doing petroleum engineering are fired when they can't project forward from the revenue based on the science of reservoir dynamics. Has any climate scientist in the history of the world ever been canned for making a bad projection?

See what happens when you make a fundamental mistake as to my specialty? I am not a geologist. You assume it...and then begin to make up stories based on your mistaken assumption, and end up in a different universe. . Pretty funny actually.

And NOW you prove you can't even read what was written. Again. I assumed nothing about error in that graph, only the normal variation that was visible within the AVERAGE range in monthly anomaly AVERAGES. This data set should CERTAINLY have been normally distributed. I provided that answer here already:

Indeed.

 

So finally, after months and months of claiming to be a scientist, Mr. Science finally admits the truth: he's really an engineer. Of course it's blindingly obvious that anyone who is incapable of citing a scientific reference, and incapable of finding a cited reference without help, cannot possibly be a real scientist. Nevertheless, it's nice to have confirmation that you've been lying to us all along.

You haven't even shown that any climate scientist in the history of the world has ever made a bad projection. And frankly, I don't think you can.

You were the one claiming to be a scientist, not me. I guess being a petroleum engineer is some branch of climatology nobody else has ever heard of.

Assume? Hell, you claimed it was, and I quote, "THE UNCERTAINTY AROUND A SINGLE POINT FOR THE ENTIRE COUNTRY". Are you now saying that you were just plain wrong when you made that all-caps shout-out? Or are you saying that there is some tangible difference between "uncertainty" and "error" that eludes the rest of the world?

That would be true only if the underlying data were stationary. Which you have not shown. If the underlying data is non-stationary, that movement will increase the naively-computed variance that you produced. And in fact, it has.

Holy buckets. You yourself are now making the exact same error that you falsely accuse climate science of making: falsely assuming a gaussian distribution. In fact, temperature data is not gaussian because of the effects of autocorrelation. Which is yet-another major Oops that you have failed to address, either in this post or the previous one.

 

My resume, as much of it as I am willing to share, is covered in one of, if not the, original post. I recommend reading and remembering what was written, you will look far less foolish every time you post something that has already been discussed, dismissed, mentioned or demonstrated.

Here...a Ternary plot of my experience.

That has been discussed in other threads. In this one we are trying to discuss the data using the language of science. Of course, this presupposes that you can actually READ of course...and as you have so ably demonstrated previously...this talent comes hard to someone of your limited abilities.

After you learn to read, we can discuss the decades required to teach you to think.

 

So ... when you claimed that you were not a geologist, in the previous post, were you lying? Or were you lying when you claimed that you were a scientist? Because as far as I can tell, the only scientific expertise you're claiming now is geology.

Which lie are you defending today, PP?

Oh what a tangled web we weave ...

I guess that's Denierspeak for, "I don't have a shred of evidence to support my claims, so I'm changing the subject."

Then perhaps you can explain to us all, in the language of science, why you think "error" and "uncertainty" are different terms. Perhaps you can explain to us all, in the language of science, why you adopted a procedure for computing the uncertainty of a single data point in the past, that gives a new and different result every time a future measurement is made.

But no, I guess you've done the smart thing and have stopped defending your indefensible math.

Wise choice. Maybe you should get a third party to QC your posts from now on.

 

I have never claimed to be a geologist. Ever. I recommend you learn to read. Again.

As much of my resume as I wish to reveal is in one of the original posts. Feel free to read it. Oops…well…learn to read FIRST..and then go and read it.

"As far as you can tell"? You have demonstrated you can't pick peer reviewed science articles to make your point when given the WORLD of literature to do so within. You can't provide a single confidence interval on any data of your own to refute a single estimate I have made to date…on ANY data…let alone the temperature station I have explicitly laid out as my example. You have proven you don't understand variance within any distribution and have pretended that I did when I did not because of that lack of understanding on your part. You don't understand uncertainty in general, and don't know the difference between uncertainty and "bad" data. You manufacture careers for others that don't exist and in general can't form a coherent thought or assemble a logical conclusion from two facts without me telling you what the answer should be in advance.

Did I miss anything on the quality of what YOU are able to TELL….anyone?

I didn't say they were. I said that YOU don't know the difference between "bad data" and uncertainty. Now you are implying that you don't know the difference (assuming there is any, depending on context) between "error" and "uncertainty"?

Google is your friend…learn to read…and then start there. If I get into issues of "bad data", or "error", I will certainly do my best to be specific…won't help you of course because you don't have the ability to think for yourself, but I do try and be clear when explaining what I am doing to those who can.

I have not yet computed the uncertainty of any single point, let alone used a method that calculates the uncertainty around said imaginary given point. Learn to read, then learn the language of science a little, and then try and ask a coherent question about what I AM doing…as opposed to this nonsense where you make up a new career for me, or claim I am doing things that I most certainly am not.

 

Arriving heat from the sun didn't change enough to account for differences we see on earth.

So, yes, the heat is going somewhere.

I don't know what you think of as a "computer model" (as there are models for pretty much everything, this being a computer age and climatology involving gigantic numbers of measurements) so I can't comment on your computer fears.

I don't know where you got your 51 number, but climatology is progressing on more fronts than 51 by a LONG ways.

Also, I'd point out that it isn't necessary for one factor to account for the missing heat. It could be that ocean cycles don't account for the full amount. Obviously, we will get improving understanding of ocean heating as studies progress.

 

What are you doing?

Everyone else is calculating an annual mean. We don't know what you're doing. Well, we kind of know, but the bigger question is why you're doing it, given that it's not relevant to calculating an annual mean.

If you could go ahead and explain your goals and process here, that'd be great, mmmkay?

 

Right now, the best way to describe what I am doing is EDA. Has everyone here seen Tukey's demonstrative graphic of Napolean's retreat from Russia? Fantastic stuff.

http://en.wikipedia.org/wiki/Exploratory_data_analysis

And if the ASA had published their approval of the aggregations, correlations, substantiated even using a mean because of an assumption of normality throughout given time periods, signed off on the confidence intervals (without or without an assumption of normality) for whatever central tendency metric was chosen...well...I could run off and read that..and THEN decide if I needed to do this much EDA to find out for myself.

Hence my high level of tolerance for the completely ignorant, such as Poor Debater. Certainly when approaching data from the perspective of EDA, the LAST thing I really want to do is to just run off and do whatever it is that others want to spoon feed me. Doesn't fit in with the spirit of EDA, scientific investigation in general, or how 3rd party independent reviewers tackle such problems. Certainly I have already explained why I would have been fired for not having these explanations sitting right beside my desk, if not already published, prior to claiming this assumption or that. In the world of science I am familiar with there is no WAY you claim something of consequence or make an assumption that has the potential to change the entire result without having it reviewed by peers, reviewed by outsiders, published it and then let people bash it with their own ideas in the literature. SOP. Right now I am testing some concepts that appeared to be central to Poor Debaters original paper and one of the references cited within it. I am also reviewing some of the concepts explained about confidence interval within the original Berkley download, going through the NOAA QC procedures on the data they have provided to see if it is necessary to go deeper into raw data because they are already building in assumptions that might effect outcome, converting the data to anomaly plots instead of absolute temperature plots just to see if they change any of the results I've generated to date, testing the TOBS adjustment (where it exists) and calculating daily mean temperatures instead of dealing with just TMax, comparing uncertainty pre-TOBS to uncertainty post-TOBS, just the basics really.

I have no doubt people want to accelerate straight to the answer. After a year of research and not having an answer, I recall myself thinking the exact same thing. However, while we all might be interested in calculating an annual mean, that is just a simple operation of adding up data (in this case NOAA QC'd data) and dividing by the number of data points. However, that alone only gives you a number. A single metric of central tendency. But that isn't the issue I am truly after. You see...if the resolution of your data is +/- 10C, and you are claiming a signal of perhaps +/- 0.3C in size...well....that is quite a problem..and until you understand your significant digits, you don't have a clue as to the value of single metric like a mean. It is just the mean.

For example...if you claim that you can find a +0.3C warming signal in the mean....and you can show your high confidence interval of +/- 0.1C around it, well goodness, that certainly we must talk because that sure looks like a verified increase to me! But if you don't calculate the uncertainties correctly, or skip them, or write a CYA assumption and assume a Gaussian with a (P5-P95)/P50 ratio of 2, proclaim they are too difficult and announce you are going to pretend they don't exist, and begin to insert assumptions instead THAT JUST HAPPEN TO NARROW YOUR CONFIDENCE INTERVALS....well....now we can discuss calling Houston and telling them we have a problem.

So sure...means are great. And I've been showing them on nearly every graph I put out. But the issue isn't just in calculating it, but determining the confidence in it. And so I have accumulated some data to see if it is even possible for a small change in mean to be statistically significant.

Sure. I have been explaining the process throughout this thread, I recommend you start reading at the beginning. To date, the most significant theoretical demonstration was my example of uncertainty around a mean given nothing more than differences in assumed correlation. Until I begin to weave together multiple stations across various levels of distance, creating perhaps a correlation surface or something akin to a semi-variogram to determine kernel and sill (perhaps some multi-variate analysis for fun?), I cannot determine the potential range of the perturbations in mean values this might induce. Could be a little. Could be alot. But I'll get there.

 

It's quite simple. His goal is to find some error that he is absolutely sure has just gotta exist, that invalidates 50 years of climate science. And since he's got a statistical package on his computer, he's pretty darn sure that he can solve the entire climate science "hoax" with a smidgen of SAS, because he's also sure that no PhD climatologist anywhere in the world has even a teeny-tiny fraction of his own superlative knowledge of the normal distribution. Heck, he even knows what AIC means, even if he'd never actually be able to find a single paper actually written by Hirotugu Akaike.

 

Translation: No idea what he is talking about so must attack.

 

It is quite simple. I am going to apply science to the problem, just as I have done before. It pains me that your lack of experience in the field blinds you to the basics of how science is done, but there is no shame in your inexperience in the field, poor reading comprehension, or lack of intellectual capability to understand it even as I explain the exact tests involved, provide the exact location where I collected the data from NOAA, or reference the pieces inside your original referenced work that make exactly the assumptions I am testing.

Just relax because, as you have so ably demonstrated, you can't participate in the conversion...not a shred of insight, not even a single different statistical test to create a proper parametric of the distributions involved, not a single confidence interval of your own, and a complete inability to even refute the ones I am providing, it is best that you stick to asking occasional inane questions proving your lack of competence..over...and over...and over.

If you need recommendations on what college courses in stats to take, to get started curing yourself of these issues, I can probably help, if you could provide the course description I would be happy to look it over and make recommendations on the ones of highest value. Can't help you as much with the ability to THINK for yourself, but unless you are really old...like over 20....then I am sure there is hope for you yet.

As far as a statistical package, please, here is one for you. But I might mention, when trying to understand a language, it isn't about having a dictionary, but in being able to use it.

http://www.r-project.org/

 

When doing science, opposing opinions aren't an attack, they are just opposing opinions. All opinions should be considered, and certainly must be considered when ones work is going through 3rd party independent review. Of course, when those opinions are accompanied by their own statistical work showing the flaws in the work being examined, those opinions carry far more weight than the nanny-nanny-booboo routine that Poor Debater specializes in. Poor Debater hasn't been able to refute a single data point, a single statistical test, presented any of his work on how all of these temperature data really are normally distributed as opposed to what I have presented, he doesn't provided a single confidence level of his own on anything, he can't defend the assumptions within the original work he referenced because...surprise....maybe they can't either...which is why they just made an assumption and kept moving along. He didn't even know such hiccups existed within peer reviewed "science"...I mean...PhD's making assumptions!! Eeeek Gads!!

As much as I rib him about not being able to think, what is really doing all my work for me, in terms of making him look bad, is just his zero experience with DOING science, as opposed to what I call "amateur-itis" where you read something about science and then pretend you now understand it. Works fine in internet forums I suppose, makes you look smart or builds your self esteem. My work had to be of sufficient quality to pass a National Academy of Science review, and did. Poor Debater is just funny as all hell in thinking that he can jawbone his way around empirical observations.

I wonder if this is a common flaw within the climate science debate? It would fit right in with the "science is settled" routine, the desire to stop questioning, particularly that related to observations. Even the averge joe can figure out that when a model says one thing, and reality another, your opinion on relative quality begins to shift.

 

I am struck, in the papers I have read and are written about in the media, the difference between what is said in the paper and what is written in the media. For instance, a recent paper someone linked to had numerous 'might be' and 'could be' in them and the determination was based on another model with uncertainties at least 100 times the rise in mm given. The paper was written to spur research but the article was a statement of fact about sea level rise. The forum member that linked it obviously thought it was fact instead of hypothesis.

 

Same here. Quick story...I was in a meeting with a reporter and me and two others were explaining, in great detail, this one little thing. We told the report, "you guys get this wrong all the time...here is how it works...now DON'T get it wrong like everyone else...please". Guess what? That same reporter, when he wrote up the meeting the next day, got exactly that one thing wrong. Enough to drive you batty. My rule now is to not talk to reporters, they write what they write, cross your fingers and hope they don't screw it up...but plan on them screwing it up.

Did they read it? My experience with Poor Debater is that he doesn't read anything except the title, and counts on his opponent to do the same. I recommend quality over quantity usually, sure you might have 50 papers ready to cut and paste titles to on a moments notice, but that has nothing to do with the ability to read ONE paper, understand the conclusions, what they are certain about, what they are not, the errors they choose to address...or skip...or CYA language to get around a particularly tricky problem. If the people doing the writing are really doing science, you can bet the farm on it. It is one of those things that publishing scientists do, including myself. But if you specialize in superficial title reading and cutting and pasting, you certainly are like a deer in the headlights the instant someone asks you a question about the article....I mean seriously...you expect them to THINK about...anything!!

 

That's nice. However, most everyone else is after an annual mean, a good simple metric for average global temperature.

So, you know you'll find something, though you're not sure what it will be. That's not really a scientific refutation of global warming science.

A red flag for pseudoscience is when someone deliberately makes the issue more complicated than it needs to be. Massage the data hard enough, and you can make it fit any outcome you want.

 

curious why it's needed?

 

Sure. But along the way to calculating that mean, there are all the assumptions, aggregations, correlations, interpolations and extrapolations that cause uncertainty all along the way of this aggregating, correlating, interpolating and extrapolating.

I have already demonstrated how the assumption of correlation alone can change the mean substantially in a simple example in my field, so now I am trying to find both the correlation matrix between stations in the a high data density area like the US, and if those correlations are themselves evident within the temperature data, and at what scale. Scale matters because of the assumed grid systems that occur later in the process of climate folks aggregating, correlating, interpolating and extrapolating temperature data. Again, I have seen the ASSUMPTIONS of this within the references provided, but amazingly not a single correlation matrix between two or more points has yet revealed itself (as opposed to how physical scientists of the USGS have provided just such a thing). Certainly I would expect that this has been done, somewhere, at some scale, it is really a prerequisite in establishing the basis for later use. Certainly I am familiar with how Hansen tried to back into large scale correlation assumptions in his "correction" of US temperature data, but that is based on something. There is no way even he would have felt safe putting that large of a bullseye on that portion of his work.

The change in mean is irrelevant if it is dwarfed by the uncertainty around that mean because of prior assumptions. Beautifully simple.

I am not trying to refute anything. I am trying to figure out if underlying assumptions comply with standard statistical analysis, or are substantiated by the underlying empirical observations.

While your experience with pseudoscience might be substantial, I have none. I am doing what I was trained to do as a scientist…ask questions of data….compare to the assumptions of others to verify it or generate my own. Upon reaching a conclusion as to underlying inconsistencies, I will be sure to provide, as I have been, both the source of the information, the way it was analyzed, any conclusions drawn (and some measure of confidence or uncertainty) and a comparison to how other folks have characterized, or not, the same data.

It really is simple, it is unfortunate if your statement about pseudo science is true, because the spirit and intent of science is quite simple really.

Agreed. Maybe you have just hit on why I can't find the published correlation matrix between all US land based temperature stations…..because that correlation strips bare all the massaging, doesn't allow added complexity, it nails down exactly the spatial relationship. I can't ever imagine why someone wouldn't provide that for everyone else to examine unless….there is some other reason why all the massaging is necessary?

 

PeakProphet wrote: “Agreed. Maybe you have just hit on why I can't find the published correlation matrix between all US land based temperature stations…..because that correlation strips bare all the massaging, doesn't allow added complexity, it nails down exactly the spatial relationship. I can't ever imagine why someone wouldn't provide that for everyone else to examine unless….there is some other reason why all the massaging is necessary?”

I see, and this some other reason would be what exactly? Why, that’s irrelevant of course! Whatever the motive, it places these scientists squarely in the crosshairs of the fossil fuel industry. If their integrity zero & additional funding their aim, wouldn’t they instead sell out to industry? Could you clarify what you mean by a correlation matrix, which institution or organization published this data & what search keys you’ve already tried? The following is something you said in your OP:

“This is a probabilistic world, and scientists not providing the proper quantification of estimates and conclusions are not doing anyone a service. This sets me diametrically opposed to the statements made by Schneider back in 1988.”

You then cited a statement allegedly made by Stephen H. Schneider, where he admittedly says that as scientists, they have to exaggerate the threat in order to get the public & officials to listen to them. An outrageous contention to be sure. Any idea why you wouldn’t have checked the validity of such an admission? In his book Science as a Contact Sport: Inside the Battle to Save Earth's Climate Schneider addresses this malicious attack:

He then cites exactly what you posted as fact. A pretty sloppy mistake, you know, from a fellow scientist like yourself.

 

I have no opinion on what conclusions or analysis place who or what in the crosshairs of whatever. Science isn't supposed to care. Unfortunately, during the process of politicizing it, for whatever reason, it is arguable that the very core of it what science is has been compromised by people trying to use it, or scientists, to "prove" what they want proved, as opposed to what IS.

Why does anyone sell out? Because they are zealots, they want money or fame, tenure or grants, the reasons are legion.

A correlation matrix at its most basic consists of a 2X2 grid, with two things to be correlated across the rows and down the columns. Each thing is correlated perfectly with itself, so the grids corresponding with item 1 would have a 1 in them by definition. Same with item #2. What remains is the correlation between item 1 and 2.

Perfect positive correlation means that as 1 goes up, 2 does as in. Perfectly negative correlation means that as 1 goes up, 2 goes down exactly as much as 1 went up. This is as about an imperfect explanation as I can do without sitting at my computer and creating the matrix to demonstrate it and making a picture, but if you are interested you can go over the general components of things like rank correlation here.

http://en.wikipedia.org/wiki/Rank_correlation

I don't understand what you mean when you say "search keys". So far I have not tested correlation among stations.

Hard to say it is a malicious attack when the author admits saying it, don't you think? Trying to spin the words to take the bite out of it is no different than a politician trying to walk back their inherent bias, racism, ignorance or factual error. One of the disadvantages of scientists wanting to play bobble head for reporters I suppose, you might let something out of the bag that you really would prefer remain in.

Certainly when someone can't effectively walk back their own words, I suppose Plan B is to pretend that using their own words is a malicious attack? Or is that just your spin on Schneider's attempted walk back?

 

It was just such questions about the data normalization in Mann 1998 that caused Professor Richard Muller to doubt the science of global warming and ultimately lead him to start the Berkeley Earth Surface Temperature project. Given that the primary purpose of BEST was to systematically address the five major concerns with global warming science in a systematic and objective manner, if there was any substance to these concerns, don't you think they would have found it?

 

[video=youtube_share;Sme8WQ4Wb5w]http://youtu.be/Sme8WQ4Wb5w[/video]

 

From Berkley Earth.

http://static.berkeleyearth.org/memos/Global-Warming-2014-Berkeley-Earth-Newsletter.pdf

 

So the scientifically accurate statement would be that 2014 is probably the warmest year on record.

In other words, Earth's average temperature for the last decade was already so high that another record warm year didn't really affect it.