Now, it certainly seems plausible that there are models out there with variables and assumptions that result in no warming, or a cooling. What is the likelihood these would get published based purely on their results?
Close to 100%. There are "fringe" journals such as the notorious Energy and Environment that are extremely friendly to critics of global warming. While not highly regarded by serious scientists, there is little doubt that E&E would publish such a model. Besides, one can publish one's models on the internet these days. Many of the models used by climate scientists are available to the public so one could get a head start by modifying an existing model. And there is little doubt that many of the fossil fuel companies would be happy to fund the development of such a model. Heck, I imagine you could get enough money to fund such a study just by asking for donations on right-wing websites. Isn't it curious that nobody has managed to produce such a model to date. Of course, maybe it isn't actually all that easy to come up with a model that is reasonably consistent with known physics, with the historical climate data, and with the climatic effects of "natural experiments" like volcanic eruptions, and yet does not predict substantial warming in response to continued CO2 emissions...
Have you ever been to Montecito? It's a lovely little town where the hills come almost down to the water. You can be pretty high up and still have an ocean view. Perhaps Gore is anticipating his house in the hills becoming beach-front property....
Remember, back in the 70s, the climate scientists were telling us all that we were going to go into a massive ice age at any minute.
Actually they weren't. It is illustrative of the level of propaganda being generated by those who hope to discredit climate science either on ideological or financial grounds that this long-debunked urban myth continues to be repeated and believed.
When you also notice that nature itself creates a lot of global warming gases when it makes volcanoes
This too is a falsehood. But like the first it continues to be repeated. This is the hallmark of crank pseudoscience, whether it be creationism or AGW denial: no false argument ever dies, no matter how conclusively it has been debunked.
Over 10% in such a short period of time? That's pretty impressive. Of course, virtually every major scientific society in the world has previously come out in support of climate science and concerns about global warming
Certainly! Just give me access to the raw, un-adjusted data that these scientists have been hoarding for decades. Oh wait, they keep destroying it.
Sorry, but somebody has been lying to you. The raw, unadjusted data is owned by various national meteorological services, and it has not been destroyed. Some of it is available for a fee, but quite a bit is available freely. You can find it here
Also, lets look at what their models from 10 years ago predicted that the weather would be for the next 10 years and compare to the historical record.
I'm asserting that they need a denser distribution than they already have
And how do you know just what distribution is required without a theory? If you don't like their conclusions, you can always insist that the conclusions are wrong because the distribution is not dense enough. If they average 400 miles apart, you can assert that they need 200. If they are 200 miles apart, you can insist that they need 100. Lacking a theory, you have no grounds to make such an assertion, however, so it is meaningless. On the other hand, climate scientists have a model that gives them a good idea of the density of measurement required to measure climate change. Now if you had an alternative model that showed that there is meaningful variation in climate change at smaller scales of distance, then you would actually have scientific grounds to make such an assertion
especially given the biases (a simple case, for the vast majority of temperature records, the sea is not included...which is 3/4 of the surface of the earth).
Do you seriously imagine that there are no measurements of sea temperatures? Or that there are not methods to correct for different densities of measurement in different parts of the world? And when you are measuring climate change, fixed biases cancel out. Climate scientists have statistical models that demonstrate that this is true. Now if you had an alternative model in which this was not the case, you might have grounds to object. But you don't, do you?
The problem is that the earth undergoes temperature changes on a dramatic basis every day - 20 degrees swing from night to day means that if you measure the temp at 2:00pm one day, and 3:00pm on another day, you have some comparability problems.
Not really. The statistical models show that this sort of thing cancels out. Now if you had a model in which this was not the case, you might have grounds to argue that they are wrong. But you don't, do you?
That's my point -> you're simply using a straw man "any other model", and asserting that any gaps in our understanding mean that your hypothesis is true. It is scientifically worthless, and a typical warmist retort.
No, I have not invoked gaps. Who is the one arguing that if there were not gaps in the density of our measurements of temperature, or in the historical climate record, then the conclusions might be different? Not me. And not climate scientists.
I'm sorry, but your explanation on that one was dubious at best. If CO2 is a positive feedback force, no matter what the trigger, it should continue to increase indefinitely.
Yes, I can understand how you might reach such a false impression, because you have no theoretical model of your own to serve as a "sanity check," so you can go on these flights of fancy, imagining that any positive feedback whatsoever, even in a system that also has negative feedbacks, will result in temperature increasing indefinitely.
Can you identify any point in history before say, 1850, where CO2 led temperature change?
Can you identify any point in history before that time in which large amounts of CO2 were so suddenly introduced into the atmosphere? The existing models reproduce the historical sequence of temperature changes and CO2 changes, as well as the modern timing. Now if you had a plausible model in which this was different, you might have credibility to argue otherwise. But you don't, do you? Indeed, nobody does. Isn't that odd? After all, there are clearly multiple wealthy commercial and national interests that stand to lose if restrictions on CO2 emissions are instituted. Surely, they have the financial resources to fund a competent scientist to develop a climate model that is consistent with the historical record and with known physics, but that doesn't predict dangerous warming in response to CO2 produced by man. Surely they would have done this, don't you think? Unless, of course, it isn't possible....
Change your resolution definition to spatial, and perhaps you can understand the issue better.
So this would be an issue if you were talking about climate models that predicted that locations a few miles apart would exhibit very different patterns of climate change. But this is not what current models predict--they are designed to predict large-scale changes, not fine spacial detail, so you don't need a dense distribution of temperature measuring stations to test them.
Further understand that there is also a temporal category - helped by automated weather stations, but unhelpful to old data captured by humans.
Temporal resolution would be an issue if climate models predicted changes on a time scale of seconds, faster than a station could record temperature change. But this also is not true--climate models predict slow, multi-decadal trend, so any thermometer has the temporal resolution to test the models
Think carefully about what you're saying here. The fact that Galileo found an alternative model that worked better does not mean that his refutation of the geocentric model could not have stood on its own. You're conflating the two inappropriately.
On the contrary, it is precisely because he compared different models that he accomplished what he did. This is true of virtually every significant scientific advance. God-of-the-gaps criticism--simply trying to find flaws or gaps in the existing model--has historically been scientifically worthless at best, and more often an impediment to genuine science
And in the case of climate science, there is no excuse for not testing your ideas by building them into a model. The fundamental physics is well known. And intuition is a poor guide when dealing with systems with positive and negative feedbacks. Constructing a concrete model is the only way to see if your ideas about how things work make sense. Witness, for example, your own confusion over the idea that CO2 can either lead or lag temperature depending upon conditions. Your intuition told you that that could not happen unless CO2 was "magic," and yet construction of a model shows that this is exactly what is expected. This is why modeling constitutes a crucial "sanity check" in thinking about such systems.
Thank you, I was asking that question earlier, and I think I understand where your misapprehension came from. Let me rephrase from the top: the surface station temperature record does not have enough resolution to assert the kinds of of trends they are asserting. This is true both in terms of area covered per station, as well as the resolution of the measurements themselves.
No, now you are using "resolution" incorrectly. I guess that I do have to explain resolution to you. Resolution describes the ability to distinguish the separation of two different objects or signals from one another. It is distinct from accuracy. For example, if the temperature were changing on a time scale of seconds, and the thermometer required minutes to reach a stable reading, then the thermometer would be said not to have the temporal resolution to measure such rapid changes in temperature. Of course, this is not the case for the measurments relevant to climate change. The trends being studied are on a much longer time scale than that required to take a temperature measurement, so resolution is not an issue. And as we have discussed, by taking multiple measurements, whether at one point in time or at many times, one has adequate precision to detect trends.
The crux of your error, though, still lies in your concept that without an alternative model, no arbitrary model can be shown to be false.
Yes, evaluation of a model with respect to alternative models remains a critical element of scientific discipline in terms of guarding against bias. Simply attacking a model and trying to find flaws is the hallmark of the crank, whether it is creationists or global warming deniers. The crank asks, "What is wrong with this model?" The scientist asks, "Which of the possible models has the greatest likelihood of producing the observed data?" Cranks attacking a scientific consensus often like to liken themselves to Galileo. But it is important to remember that Galileo did not merely attack flaws in the geocentric model--he evaluated it relative to an alternative model, the heliocentric model.
Can we also assert that given 10^14 observations of let's say, a far away star with the naked eye, we'd be able to resolve features of a 100 meters at a distance of 10 LY? Is there any limit to this statistical magic which allows us to use crude measuring instruments to resolve the finest details?
You are descending further into silliness. We were talking about the precision of measurement, not resolution.
The part where you can suddenly get more resolution out of a data set simply by looking at the sensor again.
I've already had to explain to you that precision is not the same thing as rate or accuracy. Do I now need to explain that precision is not the same thing as resolution? This is pretty basic stuff. Again, I suggest that you actually try the simple mathematical demonstrations that I described, or consult a basic statistics text.
You said earlier, "by taking a sufficient number of measurements you can get your errors down as low as you want". I'm asking you to justify that blanket assertion with an example that I believe refutes it quite cleanly, even though I take the example to an extreme to prove the point.
No. It remains possible; it's just that it would be stupid to try to do it that way. Assuming a standard deviation of a millimeter (1 million nanometers), to get a reasonable standard error of say +/- 0.1 nm, one would need to average 10^14 observations, even ignoring the obvious practicle physical difficulties of doing the comparison. So the statistical principle is valid, just not practicable in this ridiculously extreme instance. So you haven't refuted anything.
But once again, this has zero relevance to temperature measurements, where the standard deviation of measurement does not exceed the magnitude of the which you are trying to measure by 6 orders of magnitude.
You've once again added in the unjustified assumption that all random errors must have a normal distribution. What part of that aren't you getting?
Which part of "or any plausible error distribution" did you not understand?
correction, that should have been "standard deviation of thermometer measurements" in the first paragraph, not standard error. The standard error is not a fixed value, as it depends upon the number of replicates, and can be as small as you want it to be.
I'm not being clear, am I? Help me out here with the correct vocabulary -> you've got yourself a yardstick, and you want to tell if a screwdriver has increased in length by one nanometer. This isn't an offset problem, this is a problem.
Now you are just being absurd. The typical error of measurement on a yardstick is on the order of a millimeter, and you are talking about meausring something a million times smaller. Yes, it could be done in principle if you had no alternative, although you'd probably be reduced to looking your ruler and sample side by side in an electron microscope, and the number of replicate observations that you'd have to do to obtain reproducible estimates would be impractical. But what is the relevance to the topic at hand? It is not as if the standard error of thermometer measurements is +/- 10,000 degrees. It sounds to me like you are just trying to divert the discussion from your statistical misunderstandings by carrying it to a foolish extreme.
Not following you here -> you said I could choose m. Now that I've arbitrarily chosen it, why do I need to estimate it?
You take a known value of m. Then you generate simulated measurements at various times by calculating x,y pairs from the linear equation, and adding a random error (using the normal distribuition or any plausible error distribution) to each of your y values. This simulates measurments made with an instrument with some degree of statistical error. Then you do a statistical analysis (linear regression) to see how accurately and precisely you can recover your original value of m.
There is a good reason why the normal distribution is so prominent in statistics--most sources of error produce distributions that at least approximate normality in the limit. But feel free to try any plausible distribution. You will still find that the standard deviation of your means gets smaller and smaller as you take more samples. Even if you assume a highly skewed error distribution (and it would be hard to come up with a plausible explanation for why this would be so), this will only impair the accuracy of your measurments, not the precision. And an offset of the mean from the true value will not impair your ability to accurately measure trends over time.
A temperature station does not take 10 million measurements at 12:00 on the first of July - it takes one. Once you've moved past that moment, you can't take another measurement of it to improve your accuracy, even if the error distribution was normal. Take a thousand temperature stations, and a hundred years, and you still don't increase your resolution, because you're not doing your measurements at the same frozen point in time (which, essentially, is what your simple computer model will do).
Once again, your intuition is betraying you--which should be a clue that your understanding of statistics is faulty, and you should actually do the experiment before pontificating about the outcome. So try it! For example, take the equation of a line, y=mx+b, where you can chose any numbers you want for m and b. From the equation, calculate a series of y values for various values of x (corresponding to different times), do a linear regression to obtain an estimate of the slope m. Replicate the experiment a number of times for 10 x,y pairs, or 100, or 1000. Calculate the mean and standard deviation for your estimates of the trend slope for each number of samples. You will find, once again, that you can reduce the standard deviation of your mean slope arbitrarily low, coming arbitarily close to the true value, simply by increasing the number of observations over time.
Saying it again doesn't make it right. Let's apply your sad logic to your hair growth. I have a hypothesis that it grows an extra micrometer every time you say something stupid. The null hypothesis is that your hair never changes. Therefore, my hypothesis is correct, until you can build a better model.
You probably think that you are being facetious, yet you are actually on the right track. In fact, it is easy to exclude the null hypothesis that the length of my hair is constant by measuring hair length over time. So now the null hypothesis is eliminated and we have to consider different models in which hair grows. Your proposed model is indeed better than the null hypothesis, because it comes closer to the data. But of course, we have to consider other alternative hypotheses, such as constant rate growth. For example, your hypothesis predicts that my hair will not grow overnight while I am asleep. So we could measure the change in hair length overnight. This will be small relative to the precision of our measuring instrument, so we will need to do this many times in order to get the standard error down small enough to determine whether my hair length changes significantly overnight.
Think a little more carefully. You're asserting that all prior lags of CO2 to temperature increase in the historical record are because of solar variation, but that magically, after that solar variation, the added CO2 did not behave as "outside added" CO2, and cause further increase?
No, all CO2 behaves the same. Fortunately, there are other negative feedbacks in the model that prevents the system from "running away," at least with moderate increases of CO2. For example, as the average temperature increases, the amount of heat loss to space also increases, limiting further rise. This is one reason why it is important to construct mathematical models. It is very easy to mislead yourself by making hand-waving arguments about even fairly simple systems with positive and negative feedbacks. Mathematical modeling provides a crucial "sanity check" to verify that your model actually behaves in the way that you imagine.
It is worth noting that the argument you are making is an old one, debunked decades ago. Creationists do this also, making the same tired, long-debunked arguments against evolution over and over because they can't be bothered to educate themselves about what the theory of evolution actually predicts. Similarly, I expect that you will not take the trouble to actually educate yourself about climate science or statistics, and will continue to repeat the same fallacious arguments.
Um, no. Take again the "all swans are white" hypothesis. Searching for, and accumulating more and more incidences of white swan observations does nothing to make the hypothesis more likely. Searching for just one black swan, and failing miserably, is what can expand our knowledge. And all it takes is one to falsify the hypothesis.
In the perfect world of logic, perhaps. But scientists work in the messy real world. So you read a report of a black swan sighting. Does that disprove your "no black swans" theory? But wait! There are also reports of sightings of the Loch Ness monster, the abominal snowman, and space aliens of a variety of hues. Do you give those equal credence?
semantic lesson aside, you've dreadfully abused the notion of standard deviation. You've just made the ridiculous case that if you took a 640x480 pixel pictures of someone a mile away, you could figure out the length of his nose, so long as you took that picture a million times.
You are confusing yourself with a false analogy. Digital photo enhancement is a very different problem with a different set of issues. I could talk about those, but I doubt if it would make sense to somebody who has yet to grasp the concept of standard error.
Seriously, get yourself a balance scale that has a measurement resolution of 1g, and tell me how many micrograms one of your eyelashes is. Take as much time as you want.
I'm a scientist. I do this sort of thing all the time. It absolutely works. But you don't have to take my word (and the word of every basic statistics text on the planet). You could prove it yourself. Just simulate it mathematically. Take a number whose value you know, add normally distributed random noise of whatever magnitude you like. Do this several thousand times. Then try averaging 10, 100, 1000 of these numbers. Calculate the mean and standard deviation of the averages. You will find that you can recover the original number to whatever precision you like. You could even do it in Excel (here's a website that tells you how to generate normally distributed random numbers in Excel).
However, I predict that you won't bother. After all, you still haven't taken the trouble to learn about standard error. I suspect that you are only interested in science and math to the extent that you can use (or more accurately, misuse) it to buttress your preconceived notions.
Apologies to the grammar nazi for the improper use of "rate".
Your grammar is OK. It is your usage that is wrong.
So, instead of "rate", I should've used the word "accuracy".
No, you shouldn't have. You know, if you want to pretend to have expert knowledge of statistics so that you can get away with condescending remarks like "Go back to statistics 101, please," then you really should take the trouble to learn the vocabulary of the field. It kind of gives the bluff away when obviously don't know the meaning of the terms that you are throwing around.
As every undergraduate student of statistics gets drilled into them, the correct term for what you are talking about is "precision," which refers to the reproducibility of a measurement, commonly expressed mathematically as the standard deviation. "Accuracy," by the way, refers to how closely the mean of an arbitrarily large number of measurements approaches the true value. If you are trying to determine a trend, you don't care about accuracy, because the trend is unaffected by any constant bias in the measurement. For example, it doesn't matter if a thermometer is located in the middle of an asphalt parking lot and consistently reports an average temperature that is two degrees higher that one located outside of town--both will give the same trend over time.
That being said, your fantasy that enough imprecise measurements can somehow yield a very precise measurement beyond the resolution originally sampled is like those times when Chloe Obrien just needs enough time to put that grainy, pixellated image through a filter in order to generate a crystal clear picture of the bad guy in the photo
It sounds like you did not take my advice to look up "standard error" in your Stat 101 textbook (I suspect because you don't have one). That's OK, I can enlighten you.
The "standard error" is the standard deviation of a mean. The standard deviation of the mean of n observations is given by the equation,
standard error = (s.d.)/sqrt(n)
where s.d. is the standard deviation of the individual observations. It follows mathematically that the error in the mean approaches zero as the number of observations approaches infinity. In other words, you can reduce the error of your measurement to arbitrarily close to zero, no matter how poor the precision of your measuring device, merely by taking the mean of a sufficiently large number of observations. This really is (to use your own expression) "Statistics 101."
Are you high? Seriously, did you just do a big whopping hit on something? The null hypothesis of climate change is that any changes in the system do not require any human activity to drive them.
No, you don't get to decide what "null hypothesis" means. It has a well-defined statistical meaning that has nothing at all to do with human activity. The null hypothesis is that nothing is changing. Since we have overwhelming evidence that climate has changed over time, the null hypothesis has been excluded. At this point, you have to confront the question of how and why climate can change, and you are in the realm of competing models.
I don't need to have a competing hypothesis in order to falsify another hypothesis. We'll give you a simple case - I don't need to prove that some victim was murdered by someone else specifically if I can give an alibi that excludes me from being the murderer.
In science, as opposed to mathematics or logic, evidence is virtually never absolute--it is statistical. Scientific hypotheses are not proved or falsified. Rather, statistical evidence is accumulated which makes a hypothesis more or less likely. Think Bayes, not Popper. The question is what hypothesis has the greatest likelihood of producing the observed data. So it is always a matter of competing hypotheses.
No, all I need to do is prove that the model was in error. Adding ex post facto ad hoc corrections can only work for so long before you've started violating Occam's razor
Occam's Razor is another principle widely misunderstood by nonscientists. In fact, there is no scientific or logical reason to believe that a simple theory is more likely to be correct, and the historical trend has been for simple theories to be discarded in favor of more complex ones. Rather, Occam's Razor is an empirical "rule of thumb" for ordering hypotheses for investigation. For statistical reasons having to do with the number of degrees of freedom, less evidence is required to test a simple hypothesis than to test a complex one.
Are you kidding me? This is like "heads I win, tails you lose"! Please, explain under what conditions CO2 would lead a temp increase, and under what conditions CO2 would lag a temp increase
It is really quite simple. The models predict that if the increase in temperature is initiated by something other than addition of CO2 to the atmosphere--an increase in solar output, for example--then CO2 increase will lag the temperature increase. On the other hand, if CO2 is added directly to the atmosphere, then the models predict that the temperature increase will lag the CO2 increase. To put it simply, there is a positive feedback arising from the basic physics of CO2: increased atmospheric CO2 increases temperature, which reduces CO2 solubility in the ocean, which increases atmospheric CO2, which increases temperature, etc. Which comes first depends upon which starts the cycle.
The null hypothesis of climate change is that all variation is natural.
This makes no scientific sense at all. "It's not my fault" is an excuse, not a scientific model. Physics draws no distinction between "natural" and "artificial." Any scientific model of climate must depend upon hard physical parameters such as solar radiation, atmospheric gas concentrations, etc.
I've noticed that many nonscientists misunderstand the meaning of "null hypothesis." It is not some magic phrase that you can invoke to shift the burden of proof to the other guy's hypothesis. "Null hypothesis" refers exclusively to statistical measurement, and it is always the same: the null hypothesis is that nothing is changing. So the "null hypothesis" of climate change is that climate is constant and unchanging throughout time. The null hypothesis of no climate change is clearly excluded by a large body of data. Once the null hypothesis is excluded, something else does not become the null hypothesis; it is gone forever. At this point, you enter the realm of competing hypotheses, and hypotheses must be compared on equal grounds--the "burden of proof" is the same for all.
Your assertion that any area of ignorance automatically means that "man did it" is the same thing as creationists insisting that any gaps in the fossil record mean "god did it".
Please either quote the message in which I supposedly asserted that "man did it" or apologize for trying to put words in my mouth.
All I have to show is that your mathematical model, implicating man-made CO2, has errors in it that would invalidate the conclusions.
Models of complex physical systems generally have "errors" in them. For example, physical systems are often modeled Newton's "Laws" even though they have errors in them by failing to take account of relativistic effects. To show that the errors in a model invalidate the conclusions, you need to produce a corrected model without those errors and demonstrate that the conclusions are different. And to establish that your model is in fact an improvement, you need to show that it is at least as good as the original model in terms of its fit to the evidence (which for climate would be things like historical climate evidence and the climate impact of "natural experiments" such as volcanoes).
Chew on this though -> the proxy record shows CO2 concentrations lagging temp changes by 800 years, and CO2 concentrations many times higher without runaway warming. Explain how the models can be run backwards, fail to accurately depict the history we know of, yet should be trusted going forward.
You know very little about the existing models if you don't know that those models do in fact predict that depending upon conditions, CO2 can either lead or lag temperature change, and that they do in fact correctly predict the temporal order of the historical climate record.
Go back to statistics 101, please. If I've got a thermometer that has an error rate of +/- 10 degrees, I can take a million measurements and not be able to assert an accuracy of.01 degrees. Your assumption that +/- 10 degrees has some sort of bell curve that you can drive to the top of is unsupportable.
This makes no sense at all. First "+/- 10" degrees is not any kind of "rate" (here's a clue: all rates have units in terms of time, not just temperature). In statistics, a term such as "+/- 10 degrees normally refers to the standard deviation, which, contrary to your assertion, is the description of the width of a normal "bell" distribution. To understand how you can, in your terms, "drive to the top" of the bell curve by taking repeated measurements, I suggest that you pull out your "Statistics 101" text, if you have one, turn to the index, and look up the definition for "standard error" paying particular attention to the denominator.
It makes a difference when people systematically make corrections in the same direction without documentation (and then throw away the original data), delete rural stations, fail to account for urban heat islands, use defective sensor sites, use an abnormally small number of stations to estimate the temperature for vast areas, etc.
This is mere hand-waving. In fact, nobody has shown that any of these corrections make a substantive difference in the conclusions. Everybody who has tried to replicate this kind of analysis, with or without corrections, has come up with substantially the same results. Numerous peer-reviewed studies have examined this issue, and have concluded that this is not a sufficient source of bias to account for the observed trends. For a detailed review of the evidence, see The EPA's Response to Public Comments (PDF), particularly responses 2-27 through 2-39
There are always results reported that do not appear to fit with theory--any theory that you might choose. That is not the same thing as saying every imaginable contrary result has been observed.
To put in in terms of logic: "some A is B" does not imply "all B is A"
AGW critics don't need to have a falsifiable theory -> they're not making the assertion. The burden of proof here clearly belongs to the proponents of AGW.
Scientists quickly learn that you can't get away with refusing to articulate a theory of your own so that you can argue that the burden of proof is on the other guy. It is easy--and basically meaningless--to nitpick somebody else's theory if you don't have to offer an alternative. Indeed, it is virtually the hallmark of the crank. Expressing your criticisms in terms of an alternative hypothesis is a basic discipline that all scientists learn. All theories must be evaluated in the light of competing theories. Any genuine criticism implies an alternative hypothesis. If you are arguing, for example, that increased CO2 does not warm climate, you are implicitly asserting that there exists a plausible theory in which some factor other than CO2 is sufficient to explain the fact that average temperature of earth is some 30 degrees greater than would otherwise be predicted based on the Stefan-Boltzman relation. So show me the theory.
The problem that AGW folks have is that the nitpicks and flaws pointed out to them represent some very important impacts on their conclusions.
So support it with something other than handwaving. Show me an alternative mathematical theory that is equally compatible with the data.
If you have error bars of measurement that are greater than 5 degrees C, how can you predict a change of.05 degrees C?
A model predicts what the model predicts. Whether a particular measurment can be used to test the model depends upon the error of the measurement, but by taking a sufficient number of measurements you can get your errors down as low as you want. In particular, if you take multiple measurements at multiple locations or times, it can provide a very strong test of a model even if the error bars for individual measurements are large. This is basic statistics.
Look, on the one hand you want me to be skeptical of data that refutes AGW, but you won't apply the same scrutiny to data that supports AGW.
Arguing that some bit of data or some error "refutes" AGW is mere handwaving, not science. There is always error in science. The significance of error can only be evaluated in the light of a model. If you want to convince me as a scientist that you have refuted AGW, you need to do what AGW advocates have done--show me that you can produce a mathematical model of climate, based on established and testable physical principles, that does not predict global warming in response to increased CO2, and is still consistent with the existing data on historical climate change, as well as "natural experiments" such as climate perturbations in response to volcanic eruptions.
So the anti-AGW folk have it easy -> they just need to "cherry pick" data that refutes the AGW theory. Their search for data has a much, much lower bar because they don't need to have 10,000 refutations, or a million refutations, they just need one refutation. Just one bit of data that breaks the model, and the model must be changed, or abandoned.
The bigger problem of all this is that when it comes right down to it, the pro-AGW folks haven't really stated a falsifiable theory. They have in fact scrupulously avoided a falsifiable theory (warm winter? Global warming! cold winter? Global warming!), and have instead created a political movement rather than a scientific discussion.
This is immediately recongizable as the words of a scientific "back seat driver" who has never done real research, but thinks they understand science well enough to tell working scientists that they are doing it wrong. Look throught the literature on any field, and you can find data that is inconsistent with theory--whatever theory you happen to choose. Why? Because some of it is wrong. There are measurment errors, statistical anomalies, unrecognized procedural errors. If there is a theory you don't like, you can always find some data that doesn't fit. Real science owes a lot more to Bayes than to Popper. Results aren't absolute, they are statistical.
It is only the climate scientists who have a falsifiable model--mathematical models, based upon empirically testable physical processes, and making hard predictions that can be tested against historical climate data as well as modern responses to perturbations, such as volcanic eruptions. Who doesn't have a falsifiable theory? The AGW critics. Not a single one of them has managed to produce a model that is consistent with observed climate data. The pro-AGW guys have the models. The anti-AGW guys wave their hands a lot and nitpick about tiny "flaws" with no impact on the conclusions.
It shouldn't matter. If the conclusions are robust, they should not depend critically on data from any particular stations. If the conclusions change when a small fraction of the data is deleted, then something is seriously wrong. Of course, every other group that has analyzed climate data, even if the data sets were not exactly identical, has reached comparable conclusions, so it doesn't look like like of robustness is a problem here.
CRU doesn't generate raw data; all they do is analysis. It is stupid to ask CRU to retain the raw data for years, even if they had the rights to redistribute it. Scientific etiquette--and in some cases contractual restrictions--dictates that raw data should be requested from the organization that acquired it. Besides, any real scientist who wanted to check CRU's conclusions would not want old data--they'd want current, up-to-date data, so they'd have to go back to the meterological services anyway.
Slashdot Mirror
Close to 100%. There are "fringe" journals such as the notorious Energy and Environment that are extremely friendly to critics of global warming. While not highly regarded by serious scientists, there is little doubt that E&E would publish such a model. Besides, one can publish one's models on the internet these days. Many of the models used by climate scientists are available to the public so one could get a head start by modifying an existing model. And there is little doubt that many of the fossil fuel companies would be happy to fund the development of such a model. Heck, I imagine you could get enough money to fund such a study just by asking for donations on right-wing websites. Isn't it curious that nobody has managed to produce such a model to date. Of course, maybe it isn't actually all that easy to come up with a model that is reasonably consistent with known physics, with the historical climate data, and with the climatic effects of "natural experiments" like volcanic eruptions, and yet does not predict substantial warming in response to continued CO2 emissions...
Have you ever been to Montecito? It's a lovely little town where the hills come almost down to the water. You can be pretty high up and still have an ocean view. Perhaps Gore is anticipating his house in the hills becoming beach-front property....
Actually they weren't. It is illustrative of the level of propaganda being generated by those who hope to discredit climate science either on ideological or financial grounds that this long-debunked urban myth continues to be repeated and believed.
This too is a falsehood. But like the first it continues to be repeated. This is the hallmark of crank pseudoscience, whether it be creationism or AGW denial: no false argument ever dies, no matter how conclusively it has been debunked.
Over 10% in such a short period of time? That's pretty impressive. Of course, virtually every major scientific society in the world has previously come out in support of climate science and concerns about global warming
Sorry, but somebody has been lying to you. The raw, unadjusted data is owned by various national meteorological services, and it has not been destroyed. Some of it is available for a fee, but quite a bit is available freely. You can find it here
Certainly. Such a comparison may be seen here
And how do you know just what distribution is required without a theory? If you don't like their conclusions, you can always insist that the conclusions are wrong because the distribution is not dense enough. If they average 400 miles apart, you can assert that they need 200. If they are 200 miles apart, you can insist that they need 100. Lacking a theory, you have no grounds to make such an assertion, however, so it is meaningless. On the other hand, climate scientists have a model that gives them a good idea of the density of measurement required to measure climate change. Now if you had an alternative model that showed that there is meaningful variation in climate change at smaller scales of distance, then you would actually have scientific grounds to make such an assertion
Do you seriously imagine that there are no measurements of sea temperatures? Or that there are not methods to correct for different densities of measurement in different parts of the world? And when you are measuring climate change, fixed biases cancel out. Climate scientists have statistical models that demonstrate that this is true. Now if you had an alternative model in which this was not the case, you might have grounds to object. But you don't, do you?
Not really. The statistical models show that this sort of thing cancels out. Now if you had a model in which this was not the case, you might have grounds to argue that they are wrong. But you don't, do you?
No, I have not invoked gaps. Who is the one arguing that if there were not gaps in the density of our measurements of temperature, or in the historical climate record, then the conclusions might be different? Not me. And not climate scientists.
Yes, I can understand how you might reach such a false impression, because you have no theoretical model of your own to serve as a "sanity check," so you can go on these flights of fancy, imagining that any positive feedback whatsoever, even in a system that also has negative feedbacks, will result in temperature increasing indefinitely.
Can you identify any point in history before that time in which large amounts of CO2 were so suddenly introduced into the atmosphere? The existing models reproduce the historical sequence of temperature changes and CO2 changes, as well as the modern timing. Now if you had a plausible model in which this was different, you might have credibility to argue otherwise. But you don't, do you? Indeed, nobody does. Isn't that odd? After all, there are clearly multiple wealthy commercial and national interests that stand to lose if restrictions on CO2 emissions are instituted. Surely, they have the financial resources to fund a competent scientist to develop a climate model that is consistent with the historical record and with known physics, but that doesn't predict dangerous warming in response to CO2 produced by man. Surely they would have done this, don't you think? Unless, of course, it isn't possible....
So this would be an issue if you were talking about climate models that predicted that locations a few miles apart would exhibit very different patterns of climate change. But this is not what current models predict--they are designed to predict large-scale changes, not fine spacial detail, so you don't need a dense distribution of temperature measuring stations to test them.
Temporal resolution would be an issue if climate models predicted changes on a time scale of seconds, faster than a station could record temperature change. But this also is not true--climate models predict slow, multi-decadal trend, so any thermometer has the temporal resolution to test the models
On the contrary, it is precisely because he compared different models that he accomplished what he did. This is true of virtually every significant scientific advance. God-of-the-gaps criticism--simply trying to find flaws or gaps in the existing model--has historically been scientifically worthless at best, and more often an impediment to genuine science
And in the case of climate science, there is no excuse for not testing your ideas by building them into a model. The fundamental physics is well known. And intuition is a poor guide when dealing with systems with positive and negative feedbacks. Constructing a concrete model is the only way to see if your ideas about how things work make sense. Witness, for example, your own confusion over the idea that CO2 can either lead or lag temperature depending upon conditions. Your intuition told you that that could not happen unless CO2 was "magic," and yet construction of a model shows that this is exactly what is expected. This is why modeling constitutes a crucial "sanity check" in thinking about such systems.
No, now you are using "resolution" incorrectly. I guess that I do have to explain resolution to you. Resolution describes the ability to distinguish the separation of two different objects or signals from one another. It is distinct from accuracy. For example, if the temperature were changing on a time scale of seconds, and the thermometer required minutes to reach a stable reading, then the thermometer would be said not to have the temporal resolution to measure such rapid changes in temperature. Of course, this is not the case for the measurments relevant to climate change. The trends being studied are on a much longer time scale than that required to take a temperature measurement, so resolution is not an issue. And as we have discussed, by taking multiple measurements, whether at one point in time or at many times, one has adequate precision to detect trends.
Yes, evaluation of a model with respect to alternative models remains a critical element of scientific discipline in terms of guarding against bias. Simply attacking a model and trying to find flaws is the hallmark of the crank, whether it is creationists or global warming deniers. The crank asks, "What is wrong with this model?" The scientist asks, "Which of the possible models has the greatest likelihood of producing the observed data?" Cranks attacking a scientific consensus often like to liken themselves to Galileo. But it is important to remember that Galileo did not merely attack flaws in the geocentric model--he evaluated it relative to an alternative model, the heliocentric model.
You are descending further into silliness. We were talking about the precision of measurement, not resolution.
I've already had to explain to you that precision is not the same thing as rate or accuracy. Do I now need to explain that precision is not the same thing as resolution? This is pretty basic stuff. Again, I suggest that you actually try the simple mathematical demonstrations that I described, or consult a basic statistics text.
No. It remains possible; it's just that it would be stupid to try to do it that way. Assuming a standard deviation of a millimeter (1 million nanometers), to get a reasonable standard error of say +/- 0.1 nm, one would need to average 10^14 observations, even ignoring the obvious practicle physical difficulties of doing the comparison. So the statistical principle is valid, just not practicable in this ridiculously extreme instance. So you haven't refuted anything.
But once again, this has zero relevance to temperature measurements, where the standard deviation of measurement does not exceed the magnitude of the which you are trying to measure by 6 orders of magnitude.
Which part of "or any plausible error distribution" did you not understand?
correction, that should have been "standard deviation of thermometer measurements" in the first paragraph, not standard error. The standard error is not a fixed value, as it depends upon the number of replicates, and can be as small as you want it to be.
Now you are just being absurd. The typical error of measurement on a yardstick is on the order of a millimeter, and you are talking about meausring something a million times smaller. Yes, it could be done in principle if you had no alternative, although you'd probably be reduced to looking your ruler and sample side by side in an electron microscope, and the number of replicate observations that you'd have to do to obtain reproducible estimates would be impractical. But what is the relevance to the topic at hand? It is not as if the standard error of thermometer measurements is +/- 10,000 degrees. It sounds to me like you are just trying to divert the discussion from your statistical misunderstandings by carrying it to a foolish extreme.
You take a known value of m. Then you generate simulated measurements at various times by calculating x,y pairs from the linear equation, and adding a random error (using the normal distribuition or any plausible error distribution) to each of your y values. This simulates measurments made with an instrument with some degree of statistical error. Then you do a statistical analysis (linear regression) to see how accurately and precisely you can recover your original value of m.
There is a good reason why the normal distribution is so prominent in statistics--most sources of error produce distributions that at least approximate normality in the limit. But feel free to try any plausible distribution. You will still find that the standard deviation of your means gets smaller and smaller as you take more samples. Even if you assume a highly skewed error distribution (and it would be hard to come up with a plausible explanation for why this would be so), this will only impair the accuracy of your measurments, not the precision. And an offset of the mean from the true value will not impair your ability to accurately measure trends over time.
Once again, your intuition is betraying you--which should be a clue that your understanding of statistics is faulty, and you should actually do the experiment before pontificating about the outcome. So try it! For example, take the equation of a line, y=mx+b, where you can chose any numbers you want for m and b. From the equation, calculate a series of y values for various values of x (corresponding to different times), do a linear regression to obtain an estimate of the slope m. Replicate the experiment a number of times for 10 x,y pairs, or 100, or 1000. Calculate the mean and standard deviation for your estimates of the trend slope for each number of samples. You will find, once again, that you can reduce the standard deviation of your mean slope arbitrarily low, coming arbitarily close to the true value, simply by increasing the number of observations over time.
You probably think that you are being facetious, yet you are actually on the right track. In fact, it is easy to exclude the null hypothesis that the length of my hair is constant by measuring hair length over time. So now the null hypothesis is eliminated and we have to consider different models in which hair grows. Your proposed model is indeed better than the null hypothesis, because it comes closer to the data. But of course, we have to consider other alternative hypotheses, such as constant rate growth. For example, your hypothesis predicts that my hair will not grow overnight while I am asleep. So we could measure the change in hair length overnight. This will be small relative to the precision of our measuring instrument, so we will need to do this many times in order to get the standard error down small enough to determine whether my hair length changes significantly overnight.
No, all CO2 behaves the same. Fortunately, there are other negative feedbacks in the model that prevents the system from "running away," at least with moderate increases of CO2. For example, as the average temperature increases, the amount of heat loss to space also increases, limiting further rise. This is one reason why it is important to construct mathematical models. It is very easy to mislead yourself by making hand-waving arguments about even fairly simple systems with positive and negative feedbacks. Mathematical modeling provides a crucial "sanity check" to verify that your model actually behaves in the way that you imagine.
It is worth noting that the argument you are making is an old one, debunked decades ago. Creationists do this also, making the same tired, long-debunked arguments against evolution over and over because they can't be bothered to educate themselves about what the theory of evolution actually predicts. Similarly, I expect that you will not take the trouble to actually educate yourself about climate science or statistics, and will continue to repeat the same fallacious arguments.
In the perfect world of logic, perhaps. But scientists work in the messy real world. So you read a report of a black swan sighting. Does that disprove your "no black swans" theory? But wait! There are also reports of sightings of the Loch Ness monster, the abominal snowman, and space aliens of a variety of hues. Do you give those equal credence?
You are confusing yourself with a false analogy. Digital photo enhancement is a very different problem with a different set of issues. I could talk about those, but I doubt if it would make sense to somebody who has yet to grasp the concept of standard error.
I'm a scientist. I do this sort of thing all the time. It absolutely works. But you don't have to take my word (and the word of every basic statistics text on the planet). You could prove it yourself. Just simulate it mathematically. Take a number whose value you know, add normally distributed random noise of whatever magnitude you like. Do this several thousand times. Then try averaging 10, 100, 1000 of these numbers. Calculate the mean and standard deviation of the averages. You will find that you can recover the original number to whatever precision you like. You could even do it in Excel (here's a website that tells you how to generate normally distributed random numbers in Excel).
However, I predict that you won't bother. After all, you still haven't taken the trouble to learn about standard error. I suspect that you are only interested in science and math to the extent that you can use (or more accurately, misuse) it to buttress your preconceived notions.
Your grammar is OK. It is your usage that is wrong.
No, you shouldn't have. You know, if you want to pretend to have expert knowledge of statistics so that you can get away with condescending remarks like "Go back to statistics 101, please," then you really should take the trouble to learn the vocabulary of the field. It kind of gives the bluff away when obviously don't know the meaning of the terms that you are throwing around.
As every undergraduate student of statistics gets drilled into them, the correct term for what you are talking about is "precision," which refers to the reproducibility of a measurement, commonly expressed mathematically as the standard deviation. "Accuracy," by the way, refers to how closely the mean of an arbitrarily large number of measurements approaches the true value. If you are trying to determine a trend, you don't care about accuracy, because the trend is unaffected by any constant bias in the measurement. For example, it doesn't matter if a thermometer is located in the middle of an asphalt parking lot and consistently reports an average temperature that is two degrees higher that one located outside of town--both will give the same trend over time.
It sounds like you did not take my advice to look up "standard error" in your Stat 101 textbook (I suspect because you don't have one). That's OK, I can enlighten you.
The "standard error" is the standard deviation of a mean. The standard deviation of the mean of n observations is given by the equation,
standard error = (s.d.)/sqrt(n)
where s.d. is the standard deviation of the individual observations. It follows mathematically that the error in the mean approaches zero as the number of observations approaches infinity. In other words, you can reduce the error of your measurement to arbitrarily close to zero, no matter how poor the precision of your measuring device, merely by taking the mean of a sufficiently large number of observations. This really is (to use your own expression) "Statistics 101."
No, you don't get to decide what "null hypothesis" means. It has a well-defined statistical meaning that has nothing at all to do with human activity. The null hypothesis is that nothing is changing. Since we have overwhelming evidence that climate has changed over time, the null hypothesis has been excluded. At this point, you have to confront the question of how and why climate can change, and you are in the realm of competing models.
In science, as opposed to mathematics or logic, evidence is virtually never absolute--it is statistical. Scientific hypotheses are not proved or falsified. Rather, statistical evidence is accumulated which makes a hypothesis more or less likely. Think Bayes, not Popper. The question is what hypothesis has the greatest likelihood of producing the observed data. So it is always a matter of competing hypotheses.
Occam's Razor is another principle widely misunderstood by nonscientists. In fact, there is no scientific or logical reason to believe that a simple theory is more likely to be correct, and the historical trend has been for simple theories to be discarded in favor of more complex ones. Rather, Occam's Razor is an empirical "rule of thumb" for ordering hypotheses for investigation. For statistical reasons having to do with the number of degrees of freedom, less evidence is required to test a simple hypothesis than to test a complex one.
It is really quite simple. The models predict that if the increase in temperature is initiated by something other than addition of CO2 to the atmosphere--an increase in solar output, for example--then CO2 increase will lag the temperature increase. On the other hand, if CO2 is added directly to the atmosphere, then the models predict that the temperature increase will lag the CO2 increase. To put it simply, there is a positive feedback arising from the basic physics of CO2: increased atmospheric CO2 increases temperature, which reduces CO2 solubility in the ocean, which increases atmospheric CO2, which increases temperature, etc. Which comes first depends upon which starts the cycle.
This makes no scientific sense at all. "It's not my fault" is an excuse, not a scientific model. Physics draws no distinction between "natural" and "artificial." Any scientific model of climate must depend upon hard physical parameters such as solar radiation, atmospheric gas concentrations, etc.
I've noticed that many nonscientists misunderstand the meaning of "null hypothesis." It is not some magic phrase that you can invoke to shift the burden of proof to the other guy's hypothesis. "Null hypothesis" refers exclusively to statistical measurement, and it is always the same: the null hypothesis is that nothing is changing. So the "null hypothesis" of climate change is that climate is constant and unchanging throughout time. The null hypothesis of no climate change is clearly excluded by a large body of data. Once the null hypothesis is excluded, something else does not become the null hypothesis; it is gone forever. At this point, you enter the realm of competing hypotheses, and hypotheses must be compared on equal grounds--the "burden of proof" is the same for all.
Please either quote the message in which I supposedly asserted that "man did it" or apologize for trying to put words in my mouth.
Models of complex physical systems generally have "errors" in them. For example, physical systems are often modeled Newton's "Laws" even though they have errors in them by failing to take account of relativistic effects. To show that the errors in a model invalidate the conclusions, you need to produce a corrected model without those errors and demonstrate that the conclusions are different. And to establish that your model is in fact an improvement, you need to show that it is at least as good as the original model in terms of its fit to the evidence (which for climate would be things like historical climate evidence and the climate impact of "natural experiments" such as volcanoes).
You know very little about the existing models if you don't know that those models do in fact predict that depending upon conditions, CO2 can either lead or lag temperature change, and that they do in fact correctly predict the temporal order of the historical climate record.
This makes no sense at all. First "+/- 10" degrees is not any kind of "rate" (here's a clue: all rates have units in terms of time, not just temperature). In statistics, a term such as "+/- 10 degrees normally refers to the standard deviation, which, contrary to your assertion, is the description of the width of a normal "bell" distribution. To understand how you can, in your terms, "drive to the top" of the bell curve by taking repeated measurements, I suggest that you pull out your "Statistics 101" text, if you have one, turn to the index, and look up the definition for "standard error" paying particular attention to the denominator.
This is mere hand-waving. In fact, nobody has shown that any of these corrections make a substantive difference in the conclusions. Everybody who has tried to replicate this kind of analysis, with or without corrections, has come up with substantially the same results. Numerous peer-reviewed studies have examined this issue, and have concluded that this is not a sufficient source of bias to account for the observed trends. For a detailed review of the evidence, see The EPA's Response to Public Comments (PDF), particularly responses 2-27 through 2-39
There are always results reported that do not appear to fit with theory--any theory that you might choose. That is not the same thing as saying every imaginable contrary result has been observed.
To put in in terms of logic: "some A is B" does not imply "all B is A"
Scientists quickly learn that you can't get away with refusing to articulate a theory of your own so that you can argue that the burden of proof is on the other guy. It is easy--and basically meaningless--to nitpick somebody else's theory if you don't have to offer an alternative. Indeed, it is virtually the hallmark of the crank. Expressing your criticisms in terms of an alternative hypothesis is a basic discipline that all scientists learn. All theories must be evaluated in the light of competing theories. Any genuine criticism implies an alternative hypothesis. If you are arguing, for example, that increased CO2 does not warm climate, you are implicitly asserting that there exists a plausible theory in which some factor other than CO2 is sufficient to explain the fact that average temperature of earth is some 30 degrees greater than would otherwise be predicted based on the Stefan-Boltzman relation. So show me the theory.
So support it with something other than handwaving. Show me an alternative mathematical theory that is equally compatible with the data.
A model predicts what the model predicts. Whether a particular measurment can be used to test the model depends upon the error of the measurement, but by taking a sufficient number of measurements you can get your errors down as low as you want. In particular, if you take multiple measurements at multiple locations or times, it can provide a very strong test of a model even if the error bars for individual measurements are large. This is basic statistics.
Arguing that some bit of data or some error "refutes" AGW is mere handwaving, not science. There is always error in science. The significance of error can only be evaluated in the light of a model. If you want to convince me as a scientist that you have refuted AGW, you need to do what AGW advocates have done--show me that you can produce a mathematical model of climate, based on established and testable physical principles, that does not predict global warming in response to increased CO2, and is still consistent with the existing data on historical climate change, as well as "natural experiments" such as climate perturbations in response to volcanic eruptions.
This is immediately recongizable as the words of a scientific "back seat driver" who has never done real research, but thinks they understand science well enough to tell working scientists that they are doing it wrong. Look throught the literature on any field, and you can find data that is inconsistent with theory--whatever theory you happen to choose. Why? Because some of it is wrong. There are measurment errors, statistical anomalies, unrecognized procedural errors. If there is a theory you don't like, you can always find some data that doesn't fit. Real science owes a lot more to Bayes than to Popper. Results aren't absolute, they are statistical.
It is only the climate scientists who have a falsifiable model--mathematical models, based upon empirically testable physical processes, and making hard predictions that can be tested against historical climate data as well as modern responses to perturbations, such as volcanic eruptions. Who doesn't have a falsifiable theory? The AGW critics. Not a single one of them has managed to produce a model that is consistent with observed climate data. The pro-AGW guys have the models. The anti-AGW guys wave their hands a lot and nitpick about tiny "flaws" with no impact on the conclusions.
It shouldn't matter. If the conclusions are robust, they should not depend critically on data from any particular stations. If the conclusions change when a small fraction of the data is deleted, then something is seriously wrong. Of course, every other group that has analyzed climate data, even if the data sets were not exactly identical, has reached comparable conclusions, so it doesn't look like like of robustness is a problem here.
CRU doesn't generate raw data; all they do is analysis. It is stupid to ask CRU to retain the raw data for years, even if they had the rights to redistribute it. Scientific etiquette--and in some cases contractual restrictions--dictates that raw data should be requested from the organization that acquired it. Besides, any real scientist who wanted to check CRU's conclusions would not want old data--they'd want current, up-to-date data, so they'd have to go back to the meterological services anyway.
Please quote the climate change "proponents" who have allegedly said that "the earth has been cooling for the last 15 years."