How can you extrapolate anything about future climates, from a scant 150 years of data?

Question

Geologically speaking, 150 years doesn't even equal a pimple on a fly's butt.

Answer 1

I have a PhD in statistics. I have told these people time and time again that you simply can not predict out 100 years off of 150 years of the most accurate data (and this data is anything but the most accurately maintained data). If you do, your confidence intervals would be so inherently large as to be entirely useless. They make some stupid claim that they are dealing with known physical processes so they will follow a specific pattern. It seems to make no difference to them that much of the modeling is a guess and there are many unknowns in the modeling process, including whether clouds serve as a negative or positive feedback. Aside from the fact that statistics was not developed with the caveat that you get to shorten your confidence intervals for physical processes, the amount of uncertainty in the climate is large.



You can see from the stupidity of their answers that they have no idea what they are talking about when it comes to modeling. Their pretext of knowing how a jet engine works is absolutely foolish. A better analogy is saying how long a jet engine will work before failing when you have only tested 1 jet engine to failure. Your confidence intervals will be huge.



This is to be expected though. Paul's Alias has a PhD in physics, yet does not know how to do a simple linear regression. This is something that is currently being taught in high school and I have personally taught to undergraduate psychology majors. If these processes were so well known, then they would be able to predict the weather out more than 1 week, but they can't. The averaging of temps reduces the variability, but it does not change the climate modeling process into something that is not subject to rules of statistics. Their pretense that it does shows their inability to think beyond the garbage that they have eagerly fed on.



Pegminer,

IF we knew all of the variables with a high degree of accuracy, you may be able to claim determinism, but that is simply not the case and you know it. Statistics is designed to model uncertainty. There are many factors that are uncertain in climate modeling. You can not arbitrarily point to deterministic processes and pretend like this makes a difference, when the uncertainty is so large. One can make the claim of determinism on everything, in fact many have said that if you have all of the information on the initial state of the universe, you could predict everything to the color of shirt I am going to wear tomorrow, and yet we still use statistics. Why? Because of uncertainty.



Let me use your own example. That ball that you have thrown up in the air and you know all of the forces that are currently working on that ball. Can you tell me exactly where it is going to land. From your logic, you would sya so, but the truth is that there are other forces that will come into play as the ball falls back to Earth. The force of the wind acting on that ball at the time that you measured will not be the same force as a second later. Now when you tell me exactly where it will land, you will be wrong with 100% certainty. That may be the average location, but that will not be exactly where it lands. If you understand that every can be considered deterministic, but nothing is perfectly measurable, than you will know that statistics should be used, you can place a confidence interval and an area where the ball will land and you will be right 95% of the time not 0%.



I'm going to call you out on your BS. You gave a poorly worded problems and insufficient detail. You want some idiot general answer that in no way answers the question then go to someone who has no idea about statistics. You want an answer that you can use, then provide me with the details that I have asked for. Any fool can give an opinion on an ill-formed question. A professional statistician asks questions and pulls out details so that the answer that they are giving is well-informed. If you ever run across a statistician and ask them a question and they do not ask for more details, you run, because I can tell you in no uncertain terms that that person does not know what they are doing.



Further, it is not an insult to say that someone who does not know how to perform a simple linear regression indeed does not know how to perform a simple linear regression. It is, however, an insult to imply a lack of statistical knowledge, as you have, about someone that has forgotten more statistics than you have ever known.



Paul,

I do not care about your pedigree. I am sure you are a fine physicist, but if you want to talk intelligently about computer modeling you need to have a greater knowledge in stats then you have. If you want to pretend as if you undertand computer modeling better than I, you better know how to do basic simple linear regression, or I will call you out on it.



Pegminer,

I love it when people try to speak for others. You have no idea how the conversation would go between me and a statistician in the field. Your pretense of this knowledge is laughable, just as your stupid assumption that you had defined the problem very well when you hadn't even stated what you intend to do with the data. I have seen people like you come in all of the time to the stats department for help. We have to fix your problems when if you would have come to us in the first place the problems would not have been there and you could have used a simple t-test. Your arrogance is pathetically unwarranted.



And thank you for confirming two of my points.



1.) That you are talking about a book in which the conditions, initial assumptions and what you are actually trying to do with the data are discussed. I can guarantee that this discussion is far longer than the short question you posted. Once again, your arrogance in presuming that your question was well defined is unfounded.



2.) You were simply trying to test knowledge and thus lying. I knew you were a liar from the start, thanks for once again demonstrating this so that others will know.



Pegminer,

Lying some more? Cool I have come to expect it with you. Thank god none of the garbage you do actually matters. Oh no the temps will increase by <1 degree. Affect on the world is nothing. I am glad they are keeping self-righteous pompous POSes that have no business in science busy with a non-issue.



Further, you have still not addressed the issue with the fact that your pretense to determinism is false. Why? Because you can't, then only thing you have the ability to do is some vain attempt to refence some people through third hand information and then try to denigrate a person for asking clarification on a question. Your worthless shines through once again.



Virtualidiot,

I have explained this many times but I will again. It is the same reason that all of the climate models are running hot. They use the logic that correlation equals causation from the past. The truth is that they do not use the 150 years worth of data they have, but surrogates of surrogates over millions of years to create the climate models. These surrogates of surrogate only cover a very small number of the variables that would be used for climate modeling. One of these variables is the CO2 content. So they try to find the"climate sensitivity" to CO2. They pretend as if this is the same as the correlation between CO2 and temps, but it is not. The correlation is caused by two factors. 1.) CO2 is a GHG and causes warming 2.) Warming causes the ocean to not be able to hold as much CO2. Only the first factor can be used in the climate sensitivity when modeling from the past, because in the past there was no artificial CO2 being added to the environment. Consequently, the models will alwayus overestimate. Given the exponential shape of the model created, a small overestimation now will lead to large ones 100 years from now. Now one of the things that Pegminer doesn't mention because he only pulls third hand info from statisticians, is that 150 years is nowhere near enough time to model 150 years in the future. No statistician would say that it is. They would, however, use the millions of years of data from the surrogate information. Further as the scientists are traditionally the ones who do the logic and give the numbers to the stats people to analyze, very few statisticians would take the time ensure that the mistake of correlation equals causation was not made, nor would they consider it part of their job.



The funniest part of this whoel thing is that no real climatologist would say that you are using just 150 years of data. None. The morons who proclaim their great stature, do not even seem to realize this. THEY ARE USING PALEOCLIMATE DATA!!!! HELLLLOOOOO!!!



Once again, thank god they are pulling these psuedo-scientists that talk a big game but know nothing onto a fake problem. God forbid they be placed into the pharma industry. They would say a drug works for everyone because we tested it on 1 person but it is a deterministic model. The death count would be horrific.



Oh read Dana's comment kudos for actually catching this.



Pegminer,

I truly can not comprehend your worthlessness. I ask clarification questions, do not receive the answers and now you are acting like I could not answer your question. At least when you asked the question, you had enough sense not to act like a knowitall ****. Evidently the greener crowd you hang with is sucking your brains out and replacing them with arrogance.



Pegminer, Once again since you seem to be exceeedingly slow or purposefully stupid. Paul's Alias can not do linear regression. I have a PhD in stats. You cannot predict 100 years out in the future with 150 years of data, and you will not find a statistician even amongst atmospheric climate researcher who will suggest that you can. They use the paleoclimate data and only people who have little understanding of statis

Answer 2

Well, we're only trying to predict 100 years, so using a pimple to predict the next 2/3 of a pimple doesn't seem like too much of a push. We also know, very solidly, the radiation imbalance Unless you can come up with a way to reverse the sign of this imbalance, warming is going to continue.



CO2 expeller uses a lot of words dissing the statistical knowledge of AGW realists. He also makes a point of saying that he believes it is warming, and accepts the CO2 greenhouse effect, but believes that the warming will be only about 1/3 as large. Since he's using the same mechanisms, it should be a relatively simple matter for a great statistician to explain the discrepancy in conclusions. I'm curious as to why the variability should be so large that climate scientists' estimates are worthless, yet still so small that he can make his one-degree prediction.



Edit for expeller: Climate feedbacks are dependent on atmospheric and oceanic heat content, not on what is forcing that heat content. You give no mechanism for expecting the climate sensitivity to heat from this CO2 pulse to be any different than that any other warming cycle, whether from CO2 or solar. You repeat that a warmer ocean can hold less CO2 and is outgassing - but the rate constant for atmospheric partial pressure is larger than that for temperature. You have been called on this and should know better, the graph in the reference proves you wrong.

You repeat that "climate models are running hot" You have never given any data to support this, nor have you mentioned which models. As someone (over)proud of your statistical ability, it seems odd that you would not do this.

You lack both theoretical justification and empirical data.

Answer 3

If we were fitting the data to a line, and simply extending the line into the future, then yes, it would be indefensible.



However, the projections aren't just a line fit. We're not extrapolating at all. There is a thorough difference between simulation and extrapolation.



Each point in time is governed by a set of rules (physics.) In extrapolation the only thing defining each future point is a least-squares fit of the existing data. If we take the current place-and-time for a 2.5-degree (for example) grid over the planet, and define the interactions between each vertical layer of the grid cell (the layers of the atmosphere) and each neighboring grid cell by the laws of physics (fluid dynamics, thermodynamics, kinematics, etc.) then we get a deterministic model for the atmospheric system. It's inertial in nature - it depends on the boundary conditions. If two simulations begin with different boundary conditions, they can have wildly different results. However, the results are predictable based on the physics (which are known) and the boundary conditions (which are assigned probabilistically.)



The physics is (are?) understood. We don't need 150 years of record to understand the greenhouse effect. However, historical proxies enhance our understanding of the current place-and-time of our atmospheric systems. It improves our understanding of atmospheric dynamics through a large variety of inputs over time.



I view the climate models as a large Markov chain. The state of each point in time only has a memory for its prior state. It doesn't "care" what happened millions of years ago - those historical interactions are governed by physics, just like each link in the chain. It takes the current state, considers the physics of the situation, and moves forward to the next step, numerically completing the scores of physical equations that govern the interactions, reaching a deterministic result, and beginning the process again at the next step.



If you connect the dots and draw a line, you're extrapolating. You're ignoring cause and effect. You're confusing correlation with causation.



When you take the physics into account it results in a simulation, a model, not a line fit.

Answer 4

Pegminer is exactly right, it's because future projections are based on physics. We use past data to make the models and future predictions as accurate as possible - a process called 'hindcasting' - but future projections are based on physics.



We have paleoclimate data going back tens of millions of years, by the way. Of course the further back the data goes, the larger the error bars, but pretending this data doesn't exist is simply denial. Ice core data goes back nearly a million years and is very useful.

Answer 5

The Moberg Graph cited here by 'David', “UAH Lower Troposphere Annual Average Anomaly” is not about warming in the whole atmosphere, but about temperature changes in one layer of the atmosphere, the troposphere, closest to Earth and with most of the clouds, and which shades and cools the Earth as increased heat causes extra evaporation and therefore more clouds.

Where does the increased heat come from? It comes from the stratosphere, the air layer above the troposphere, where there are only thin wispy clouds, which warm Earth more than the troposphere cools it, with infrared radiation, increased by greenhouse gases, mainly carbon dioxide.



In Science as a Contact Sport Stephen Schneider explains that this mistake was the basis for his famous prediction in the 1970s, that the Earth may be cooling.(made before computer model were possible.)However, soon after, he corrected his mistakes, explaining that the Earth shows a warming trend when you include measurements from the stratosphere.



“Science,” he said, is “continuously correcting its conclusions based on new research”. We cannot perform experiments on the world’s climate, but we can make estimated forecasts, check their accuracy after a period of time, add information from new research, and then make new and more accurate forecasts.



By 1980, computers were able to handle large amounts of information, and make more and more accurate predictions, showing, unfortunately, that global warming is happening faster and with more intensity than anyone imagined.



I am sure that the many of the PR people for the American Petroleum Institute feel bad now about their work on a campaign to “recruit a cadre of scientists … to help convince journalists, politicians and the public that the risk of global warming is too uncertain to justify controls on greenhouse gases.” In the mid 1990s, when they started this, the science was not as clear, and the urgency of the need to stop global warming not widely understood.

Answer 6

You also have to remember that it has been only about 40 years when we have been getting better information from most of the Earth. We still haven't been monitoring the oceans for that long. But according to the climate scientists ice cores at the poles are more accurate then past vegetation growth found in bogs and swamps through out the World. Then, these same scientists cherry pick their average period to co-inside with the cool down during the mid 1900's.



Wow, got to love the people who think that simple physics will solve the complex model of the Earth especially since most climate scientists are more worried about proofing man made climate change so that they can keep their jobs. All you have to do is look at the number of retired climate scientists who have said that they were forced to say what they did.

Answer 7

Geologically speaking, we're also not predicting the climate of the next Epoch.



The data that we have spans back millions of years. We can see patterns in the way climate acts; we can see feedback loops, the correlations between Milankovitch cycles and climatic cycles, when temperatures rose or fell, when CO2 levels were high or low, so on. We can use this information combined with existing trends to project upon the next century what will happen if we follow certain lines of action. The theory, as Paul's Alias 2 nicely elaborated for us, allows us to explain data and then predict future data.





Funny thing David about plotting two curves on the same graph is that you have to change the axes in order to make them fit. Of course there is no noticeable difference between the temperature plots of the 1912-1945 and 1975-2009 periods, or the Medieval Warming period plotted by Moberg and the CRU current temperatures - except that the plots in each are off by 0.2˚C. They're really indistinguishable? Nothing different between them?



Why don't you also graph the data available so far from 2010 onto the HadCRU graph? Is it because the curve would take a deviation from the Medieval Warming Period since it would put the curve above 0.7˚C? Hm?



You're one to talk about cherry-picking. Not to mention simply lying with the graphs.

Answer 8

You can't.



One of the “problems” with the way climate data are handled is in the obsession with applying linear trend lines to non-linear data.



A sine wave has no linear trend…



http://www.woodfortrees.org/graph/sine:10



But… What happens if my data represent only a portion of a Sin wave pattern?



http://i90.photobucket.com/albums/k247/dhm1353/Climate%20Change/sinwave.png



The r-squared of a linear trend line of this partial sine wave is 0.88… 88% of the data fit the trend line. This implies a very strong secular trend; yet, we know that in reality sine waves do not have secular trends.



If we take the entire HadCRUT3 series and apply a linear trend line, we get an apparent secular trend…



http://i90.photobucket.com/albums/k247/dhm1353/Climate%20Change/HadCRUT3.png



The r-squared is 0.55… 55% of the data fit the secular trend. This implies that there is a real long-term warming trend.



What happens to that secular trend if we expand our time series like we did with the sine wave?



The apparent secular trend vanishes in a puff of mathematics…



http://i90.photobucket.com/albums/k247/dhm1353/Moberg-1.png



How can such a clear secular trend vanish like that? The answer is easy. Each “up hill” and each “down hill” leg of a Sine wave has a very strong secular trend. Unless you have enough data to see several cycles, you don’t know if you are looking at a long-term trend or an incomplete cycle.



If we take the HadCRUT3 series and compare the the period from 1912-1945 to the period from 1975-2009, we find that they are statistically indistinguishable…



http://i90.photobucket.com/albums/k247/dhm1353/Climate%20Change/Hadley.png



We also find that Moberg’s Medieval Warm Period reconstruction is very similar to the HadCRUT3 series…



http://i90.photobucket.com/albums/k247/dhm1353/Climate%20Change/MobergvHadCRUT3.png



What does all of this mean?



It means that the Earth’s climate is cyclical. It means that the climate changes we’ve experienced over the last 150 years are not anomalous in any way, shape, fashion or form.



The so-called consensus is living in a Goldilocks sort of world. Geological-scale data are too long (misleading) and any time series short enough to not show a warming trend is too short (cherry picking). The instrumental record is "just right." Since the "Hockey Team" climate reconstructions fit into the Goldilocks world, they are sound. Other reconstructions (Esper, Moberg, Loehle) are treated almost like heresy... Because they don't conform to the Hockey Team.

Answer 9

Like Paul's Alias started with, the logic's the same.



We've only had jet engines for about 70 years or so. Yet I'm pretty confident we know how they work and can use this to predict what ones will work in the future - that's the wonderful thing about science.





150 years is enough time, if we had good data, to determine feedbacks that are relevant on a decade or century timescale - which is what we're interested in.

Answer 10

<>



Your question makes no sense. We've had less than 70 years of knowledge about nuclear weapons. Does that mean that we have no basis to think a wide-scale nuclear war would not have catastrophic effects.



Often in science one does not really need any data to have great confidence in a theoretical result. In high school they teach you about the "scientific method" but that us really just something high school teachers believe religiously and real scientists often ignore. For example, both Special Relativity and General Relativity were widely believed with ZERO years of data, not your "150 year" requirement. The existence of anti-particles was predicted, and accepted by serious scientists (but maybe not "scientific method" non-scientists) before any anti-particle was seen. We know where in the sky Mars will be 150 years from now just onthe basis of theory.



Let me give you another example of the power of theory. If someone were to ask "If Matilda has 874 apples and a 98 year old Armenian you used to be a champion soccer player gave her 67 apples how many apples would she have?" I could get the result without finding someone named Matilda, without finding a 98 year old Armenian guy who used to be a champion soccer player and without handing them apples to do the experiment. That is why science is so much more powerful than the "scientific method" (sic).



Global warming can be expected theoretically because we know that objects emit a blackbody spectrum and we know CO2 has a resonance at a major frequency of that spectrum. We do not need to get actual data to see that the system will hea up. There is some use for actual data because there are complex feedback effects, but non-scientists like yourself do not understand that science is not sitting around collecting piles of data and using slide-rules.



EDIT



<>



That is why it is so hard to believe you really have a PhD in statistics.



Why not tell me how many years you do need,and derive it mathematically since you think there is some special number.



EDIT



<>



Perhaps I should have spent more time in high school and not have gone to MIT undergrad and University of Texas and University of California at Santa Barbara for grad school. There is no required course at MIT to get an undergrad physics degree nor anyrequired course at UT or UCSB to get a PhD in physics to learn linear regression. I also never took accounting.



EDIT



<>



You are failing to understand the concept of "science". Science allows one to make accurate predictions of behaviors without necessarily testing them out first with years of data. One does not need billions of years of data to have a good understanding of what added CO2 would do to the Earth's temperatures.

Answer 11

Easy, it's called PHYSICS. If I know the position, velocity and forces on a ball I can extrapolate its future position without giving a hoot what it was doing 5 minutes ago, much less 150 years ago. Of course climate is more complicated than the trajectory of a ball, but the principles involved are the same. People have been working on numerical prediction of atmospheric motions since Lewis Fry Richardson did it while driving ambulances in World War I. That's nearly one hundred years of research on the subject, with vast improvements in input data and computing power coming in the last 20 years.



"Weather Prediction by Numerical Process" by Lewis Fry Richardson (1922)

"Introduction To Three-dimensional Climate Modeling" by Warren Washington and Claire Parkinson (2005)



EDIT: I expel... makes the mistake of thinking that climate forecasting is ONLY done with statistical models. What he says might be true if we were using purely statistical models of forecasting, but we are not. Climate is to a very large extent deterministic, which makes his argument baseless.



I expel... should also refrain from insulting others' knowledge of statistics. I recently asked a statistics question on here and failed to receive any sort of useful answer from him, but did receive useful information from someone involved in climate science, who seemed to have a much better grasp of this branch of statistics. I have no reason to doubt that I expel... has a Ph.D. in statistics, but I also have good reason to doubt that he has little knowledge of the field as applied to weather and climate.



https://answersrip.com/question/index?qid=20100713091217AAMEHIM



Another EDIT: I find it continually astonishing that "denialists" assume incompetence of everyone that works in the climate field. The denialist geologist with the Bachelor's degree thinks he knows more about geology and climate than the Ph.D. geologist at Columbia University that works in the field and has many honors. The software engineer denialist believes that software engineering principles are not applied in the climate field, despite the many software engineers working on climate models (probably from better schools and with more degrees than he has), and of course the resident Ph.D. statistician denialist thinks he knows more about the statistics of weather and climate than do the Ph.D. statisticians that actually work in the field. Get over yourselves people! Despite the fact that I expel... could give no answer to what was actually a quite well-defined problem in time series as applied to meteorological data, he somehow he thinks that he has some sort of insight that statisticians working in field of climate don't have. Everyone knows about what not having perfectly accurate initial conditions does--Lorenz was a climate scientist, after all (not a statistician). That problem is addressed in climate models--pick up a book on the subject, you might learn something. I'd start with "Atmospheric Modeling, Data Assimilation and Predictability" by Eugenia Kalnay. You might also want to pick up some books on statistics as applied to atmospheric science, since you seem woefully ignorant of the subject, but happy to inject your lack of knowledge into any discussion.



I expel... is completely losing touch with reality, and more than a little self-centered. He says



"You were simply trying to test knowledge and thus lying. I knew you were a liar from the start, thanks for once again demonstrating this so that others will know."



He says this because he thinks I only asked the question to "test" him, when actually I needed the information for a climate paper that I'm currently writing and I learned much valuable information both on and offline from others. I'm sure if he'd been able to answer the question he would have crowed about it, but he couldn't and now he claims the question was not real. Again, get over yourself, the world does not revolve around you, and there are lots of competent Ph.D. statisticians that actually DO work on climate and understand the issues.



Final EDIT: Sorry to sound like a knowitall I expel... I guess I shouldn't have started off my answer telling everyone that I had a Ph.D. in statistics, implying that no one in climate science knows statistics and that Paul's Alias can't do linear regression.....oh wait, that wasn't me, was it?