According to the Kasting book I cited in my other comment, turning up insolation by 10% would cause the tropopause to greatly expand (from ~15 to ~150 km), allowing more water to escape into the stratosphere instead of being trapped in the troposphere. Once it's in the stratosphere, photodissociation can occur (UV splits apart H2O) since the atmosphere isn't screening UV up there, and the H2 preferentially escapes over the O2. The planet loses all its water over time.
Once a planet loses its liquid water, silicate weathering shuts down. Silicate weathering is the main geochemical sink of atmospheric CO2. Over time, the solid planet eventually outgasses all its CO2 (assuming it has active volcanism). All the CO2 ends up in the atmosphere, and nothing exists to take it back out again. There's no carbon cycle. That's what likely happened to Venus.
There's something wrong with an estimate that puts Venus (at 0.723 AU) inside the habitable zone, since it has a runaway greenhouse. That's a pretty good working definition of "uninhabitable". The 0.725 bound comes from an early paper with a simplified model. The later papers listed on Wikipedia (Hart et al., Fogg, Kasting et al.) all put the inner edge at around 0.95 AU, which we're even closer to. But there is some debate as to what would be "uninhabitable". According to Kasting, at 0.95 AU you wouldn't have a runaway greenhouse like Venus, but you would eventually lose all the planet's water.
Venus is well outside the habitable zone, for obvious reasons. It's not near habitability. If you moved the Earth inward from 1 AU to 0.95 AU, the stratosphere would moisten and you'd gradually lose all the planet's water to photodissociation followed by hydrogen escape. This is arguably the inner edge of the habitable zone. If you moved the Earth in to 0.85 AU, you'd boil the oceans and produce a runaway greenhouse. Venus is at about 0.72 AU.
I've just sucked one year of your life away. I might one day go as high as five, but I really don't know what that would do to you. So, let's just start with what we have. What did this do to you? Tell me. And remember, this is for posterity so be honest. How do you feel?
Re:Isn't AGW supposed to be about _rate_ of change
on
Bastardi's Wager
·
· Score: 1
Yes, the rate of change is the more important concept. After all, if temperatures stayed constant but higher than the 20th century average, maybe you could say "Hey, the planet is hotter than it's ever been", but it still wouldn't be warming. There is a minor amateur blog industry involved in comparing climate model predicted warming rates to observed warming rates.
Frankly, this is very difficult to do right. To compare trends you have to be careful about statistical significance (are the predicted and observed trends "significantly" different), which means getting the error bars on the trends right. This in turn means having an appropriate statistical model of random temperature fluctuations. And if you're looking at short term trends, the results are sensitive to what you pick for the starting year of the trend (does it happen to be unusually warm or cold). There are also questions of how to deal with climate model uncertainty: if different models make predictions of different rates, how do you judge the general question of "what climate science predicts". This has to involve comparing models, some of different quality.
Even if you get past the statistical problems, there are other problems involving the assumptions made to produce the predictions. If you make a prediction with a model you have to predict the warming agents right (how much CO2 and other greenhouse gases will there be in the future). How much of the error in model is attributable to climate prediction error, and how much to greenhouse gas prediction errors? Also, short-term predictions are very vulnerable to errors in the initial state. Climate models are usually not initialized or calibrated to the exact current state of the atmosphere-ocean circulation. Over the long term this doesn't matter too much, as dependence on the initial state averages out, but on decadal scale, it matters.
Long story short, you're asking the right question, but I don't know a good answer if you're asking about something as short term as, say, a 10-year warming rate.
If it's really deep, they could name it after the nameless things: "Far, far below the deepest delvings of the Dwarves, the world is gnawed by nameless things". Except they're nameless.
I can't believe how many posters missed your point. "Ha, you moron, exponential growth can't continue forever!" Yeah, that's exactly the point. Re-read the comment.
No-one would take [inflation] seriously, except that when you calculate things from it, it works incredibly well - it's the source of http://xkcd.com/54/
Not quite. You don't need inflation to get the blackbody spectrum of the cosmic microwave background radiation (CMBR) observed by the COBE satellite, which is what the xkcd comic depcits. That's a prediction of plain vanilla Big Bang cosmology, with or without an early inflationary phase.
However, inflation does predict details in the CMBR angular power spectrum, the "acoustic peaks", which were observed by the later WMAP satellite. And it solves other "paradoxes", like the horizon problem you mention.
Yes, we know that reduced order models already exist. The new thing here is that you can input the parameter range you want, wait a few hours, and the supercomputer sends you back a custom reduced order model optimized for the parameter range you care about. You can then apply this model "in the field" to the situation you're dealing with. It's supposed to be useful for situations where the details of the problem aren't known ahead of time, and you can't pre-compute the reduced order model.
And no, the article doesn't have details on how to calculate error bounds, because it's just a news article. The guy has been publishing on this algorithm for years. I'm sure you can look up how the errors are calculated if you want.
I just re-read the article. What it sounds like they're doing is having the supercomputer craft a reduced order model which is optimized to a particular range of parameters. That suggests to me that they're constructing a perturbative model about some fixed solution that the supercomputer produces. Perturbative approximations are more accurate the closer they are to the "reference" solution. So the innovation appears to be: the user can specify what set of parameters they want to perturb about, and therefore construct a custom model which is optimized to perform well in the parameter range that user is interested in.
What the phone is doing is "reduced order modeling", which (if the article is using the term accurately) finding a simple set of equations whose solution provably approximates a far more complex system of equations. It's not just interpolation (lookup tables, or machine learning from a training ensemble). Reduced order models actually have dynamics in them.
The potential innovation here, as I see it, is that it takes some supercomputing effort to build the reduced order model for a specific problem. So you can input a description of your problem into the phone, communicate it to the supercomputer and let it analyze the problem for a couple hours, and it spits out a custom approximation to your problem that you can run in the field on your phone.
Now, what I'm not sure that they've developed is some convenient way to specify the problem using the phone's interface. Maybe the problems they're working with are too complex to be easily specified with a few keystrokes. If all they're doing is loading a precomputed reduced model onto the phone, and there's no interaction with the supercomputer to let it handcraft solutions to problems in semi-realtime, then it's not nearly as interesting.
Yeah, fine, we'd all like to compute our likelihoods faster, and you can imagine hardware which does it, but it's not clear from the article why doing this analog is superior. Or even how you can avoid sampling algorithms like MCMC using this approach. How does making probabilities analog get you a joint probability distribution over a parameter space, let you marginalize parameters to a lower dimensional distribution, and all the other things MCMC is for? I have a feeling that this is chip is targeted at Bayesian networks, where you're just propagating conditional probability values down a graph, rather than at MCMC.
"Since it is generally impossible to measure what is important, bureaucrats instead turn their energies toward making important what is measurable." — J.M.W. Slack, Egg and Ego
Your position is utterly absurd - it seems to ignore the fact that people can 1) lie and 2) be mistaken.
I didn't say that people can't lie or be mistaken. I explicitly said that people can be mistaken. I also said that is not an excuse to discard eyewitness testimony. It is the judgment of the court whether the testimony provides a reasonable standard of evidence or not. It is your position that is utterly absurd: that no evidence such evidence can ever decide a case.
A single person saying X should not be enough for a court to decide X with NO OTHER INFORMATION WHATSOEVER.
It can be and often is. As I said, criminal trials have been decided on the basis of eyewitness testimony before, and this isn't even a criminal trial. It's up to the judge to decide whether the witness is credible and the testimony likely to be reliable; if so, the testimony can decide the case. If there is some reason to doubt the witnesses credibility -- and there could be, but this is not automatically the case -- then the testimony can be ignored.
It may not be hard to ascertain someone is going 20 mph over the speed limit, but it should not PROVE IT BEYOND A REASONABLE DOUBT.
Yes, it can prove it beyond a reasonable doubt, unless reasonable doubts have been raised about the credibility of the witness. And this isn't a murder trial. As another poster pointed out, minor moving violations don't have to be proven "beyond a reasonable doubt" in the first place, in most courts of law -- they have a lower standard of evidence.
You wouldn't believe conclusively in Aliens if a guy in Iowa said that he saw a UFO out in his field.
Look, you would come across as a lot less stupid if you'd drop these moronic analogies and just argue the case at hand. If you want to argue that there's reason to believe the cop was lying, or unable to adequately judge the speed, then argue that. But a cop visually estimating the physical speed of a vehicle is not even remotely equivalent to a random guy in NYC claiming he saw Jesus.
It's true that there can be unconscious biases in any subjective judgement which renders the judgment uncertain -- that's why it's "subjective". It's always going to come down to details. If it's 20 mph over, then it's fairly certain that you can tell that pretty reliably, regardless of what kind of car they're driving. If it's 5 mph over, then "red-car bias" could play a role, and that's a matter for the judge to decide. I'm just saying that "subjective judgement" isn't the same thing as "no evidence", as the original poster claimed.
Eyewitness testimony certainly can be, and often has been the basis of conviction, even for captial murder cases. If one guy stabs another guy right in front of you, it's not really difficult to ascertain that actually happened and wasn't a hallucination, provided the light is good, you got a good look at the guy or know him, etc.
Likewise, it's not hard to ascertain that someone is, say, going 20 mph over the speed limit just by looking.
Your attempt to conflate "he's was obviously going way over the speed limit" with "it appeared he turned into a flaming dragon" only shows how absurd your position actually is. In the real world, people are able to judge physical events which happen in front of them, such as distances, speeds, or actions.
Certainly this is not perfect -- a person can't tell reliably if someone is 1 mph over the speed limit -- and in the case discussed here, there is room for debate.
But that doesn't change the fact that it is totally ridiculous for you to claim that eyewitness testimony about physical events doesn't constitute evidence or that it's equivalent to belief in the Second Coming. Just give up and stop embarrassing yourself. You're replacing a bad analogy with a ludicrous one.
"Beyond all reasonable doubt" is only required of criminal cases, which this is not, as another poster pointed out. For a moving violation such as speeding, many courts only require a lower standard, such as "clear and convincing evidence". So, again, the question is whether subjective evidence meets this standard, in this case.
It's certainly possible to eyeball a car and, say, tell it's going 20 mph over the limit. In this particular case, the details are more ambiguous. My point is not that eyewitness testimony rates an automatic conviction. It's merely that eyewitness testimony can be sufficient for conviction, depending on circumstances, and that relying on eyewitness testimony is not equivalent to convicting somebody because he looks guilty, ignoring facts and evidence, or any of the ridiculous hyperbole of the original poster.
Of course subjective opinion is not as good as objective measurement. That's not the question. The question is whether subjective evidence, in this case, is sufficient to reasonably ascertain guilt.
To return to the previous poster's murder analogy theme: photographic evidence of one person stabbing another is certainly better than eyewitness testimony. Eyewitnesses can be mistaken. However, in the absence of someone happening to catch the murder on camera, eyewitness testimony itself may be sufficient to convict. It depends on the credibility of the eyewitness as unbiased and an assessment of whether the judgment of the witness is likely to be reliable (e.g., what were the lighting conditions, did the eyewitness have a good line of sight, etc.) That's what needs to be discussed here, not whether subjective judgement is equal to a radar gun.
That's not the spirit of it at all. The ruling is that eyewitness testimony can constitute reasonable evidence, albeit evidence which is not as strong as, say, photographic proof. It's bizarre that you can't understand this. Are you also railing against eyewitness testimony in murder trials? The real issue here is not that "appearances trump facts". The real question is whether eyewitness testimony about vehicle speeds constitutes sufficient evidence. This can be debated reasonably without ridiculous claims that "appearances are more important than facts".
It's not analogous to condemning a person for "looking wrong". It's eyewitness testimony as evidence of a person's actions: "It looked like you were speeding" is analogous to "It looked like you stabbed that guy". Yes, eyewitness judgment can be wrong, but eyewitness judgment is not the same as "you look evil therefore you are guilty".
"You look like a murderer" is more analogous to "you look like a speeder". It is quite different from "it looked like you were speeding", and has nothing to do with the case being discussed here.
Be careful with DIY ECG/EKG. You don't want to mess up and accidentally run too much current through your heart. Be sure that you use an optical decoupler to isolate the power source from the detector. (The way this works, IIRC, is you turn the electrical signal into light using an LED, then use a photodetector to convert that back into electricity, so there is no direct path of conduction between your heart and the ECG.)
Slashdot Mirror
According to the Kasting book I cited in my other comment, turning up insolation by 10% would cause the tropopause to greatly expand (from ~15 to ~150 km), allowing more water to escape into the stratosphere instead of being trapped in the troposphere. Once it's in the stratosphere, photodissociation can occur (UV splits apart H2O) since the atmosphere isn't screening UV up there, and the H2 preferentially escapes over the O2. The planet loses all its water over time.
Once a planet loses its liquid water, silicate weathering shuts down. Silicate weathering is the main geochemical sink of atmospheric CO2. Over time, the solid planet eventually outgasses all its CO2 (assuming it has active volcanism). All the CO2 ends up in the atmosphere, and nothing exists to take it back out again. There's no carbon cycle. That's what likely happened to Venus.
There's something wrong with an estimate that puts Venus (at 0.723 AU) inside the habitable zone, since it has a runaway greenhouse. That's a pretty good working definition of "uninhabitable". The 0.725 bound comes from an early paper with a simplified model. The later papers listed on Wikipedia (Hart et al., Fogg, Kasting et al.) all put the inner edge at around 0.95 AU, which we're even closer to. But there is some debate as to what would be "uninhabitable". According to Kasting, at 0.95 AU you wouldn't have a runaway greenhouse like Venus, but you would eventually lose all the planet's water.
Venus is well outside the habitable zone, for obvious reasons. It's not near habitability. If you moved the Earth inward from 1 AU to 0.95 AU, the stratosphere would moisten and you'd gradually lose all the planet's water to photodissociation followed by hydrogen escape. This is arguably the inner edge of the habitable zone. If you moved the Earth in to 0.85 AU, you'd boil the oceans and produce a runaway greenhouse. Venus is at about 0.72 AU.
See chapter 6 of this book, partly based on this paper (PDF).
I've just sucked one year of your life away. I might one day go as high as five, but I really don't know what that would do to you. So, let's just start with what we have. What did this do to you? Tell me. And remember, this is for posterity so be honest. How do you feel?
Yes, the rate of change is the more important concept. After all, if temperatures stayed constant but higher than the 20th century average, maybe you could say "Hey, the planet is hotter than it's ever been", but it still wouldn't be warming. There is a minor amateur blog industry involved in comparing climate model predicted warming rates to observed warming rates.
Frankly, this is very difficult to do right. To compare trends you have to be careful about statistical significance (are the predicted and observed trends "significantly" different), which means getting the error bars on the trends right. This in turn means having an appropriate statistical model of random temperature fluctuations. And if you're looking at short term trends, the results are sensitive to what you pick for the starting year of the trend (does it happen to be unusually warm or cold). There are also questions of how to deal with climate model uncertainty: if different models make predictions of different rates, how do you judge the general question of "what climate science predicts". This has to involve comparing models, some of different quality.
Even if you get past the statistical problems, there are other problems involving the assumptions made to produce the predictions. If you make a prediction with a model you have to predict the warming agents right (how much CO2 and other greenhouse gases will there be in the future). How much of the error in model is attributable to climate prediction error, and how much to greenhouse gas prediction errors? Also, short-term predictions are very vulnerable to errors in the initial state. Climate models are usually not initialized or calibrated to the exact current state of the atmosphere-ocean circulation. Over the long term this doesn't matter too much, as dependence on the initial state averages out, but on decadal scale, it matters.
Long story short, you're asking the right question, but I don't know a good answer if you're asking about something as short term as, say, a 10-year warming rate.
If it's really deep, they could name it after the nameless things: "Far, far below the deepest delvings of the Dwarves, the world is gnawed by nameless things". Except they're nameless.
Is it an antimatter planet traveling at a peculiarly high velocity?
I can't believe how many posters missed your point. "Ha, you moron, exponential growth can't continue forever!" Yeah, that's exactly the point. Re-read the comment.
No-one would take [inflation] seriously, except that when you calculate things from it, it works incredibly well - it's the source of http://xkcd.com/54/
Not quite. You don't need inflation to get the blackbody spectrum of the cosmic microwave background radiation (CMBR) observed by the COBE satellite, which is what the xkcd comic depcits. That's a prediction of plain vanilla Big Bang cosmology, with or without an early inflationary phase.
However, inflation does predict details in the CMBR angular power spectrum, the "acoustic peaks", which were observed by the later WMAP satellite. And it solves other "paradoxes", like the horizon problem you mention.
Yes, we know that reduced order models already exist. The new thing here is that you can input the parameter range you want, wait a few hours, and the supercomputer sends you back a custom reduced order model optimized for the parameter range you care about. You can then apply this model "in the field" to the situation you're dealing with. It's supposed to be useful for situations where the details of the problem aren't known ahead of time, and you can't pre-compute the reduced order model.
And no, the article doesn't have details on how to calculate error bounds, because it's just a news article. The guy has been publishing on this algorithm for years. I'm sure you can look up how the errors are calculated if you want.
I just re-read the article. What it sounds like they're doing is having the supercomputer craft a reduced order model which is optimized to a particular range of parameters. That suggests to me that they're constructing a perturbative model about some fixed solution that the supercomputer produces. Perturbative approximations are more accurate the closer they are to the "reference" solution. So the innovation appears to be: the user can specify what set of parameters they want to perturb about, and therefore construct a custom model which is optimized to perform well in the parameter range that user is interested in.
What the phone is doing is "reduced order modeling", which (if the article is using the term accurately) finding a simple set of equations whose solution provably approximates a far more complex system of equations. It's not just interpolation (lookup tables, or machine learning from a training ensemble). Reduced order models actually have dynamics in them.
The potential innovation here, as I see it, is that it takes some supercomputing effort to build the reduced order model for a specific problem. So you can input a description of your problem into the phone, communicate it to the supercomputer and let it analyze the problem for a couple hours, and it spits out a custom approximation to your problem that you can run in the field on your phone.
Now, what I'm not sure that they've developed is some convenient way to specify the problem using the phone's interface. Maybe the problems they're working with are too complex to be easily specified with a few keystrokes. If all they're doing is loading a precomputed reduced model onto the phone, and there's no interaction with the supercomputer to let it handcraft solutions to problems in semi-realtime, then it's not nearly as interesting.
Not really. In this case, the smart phone isn't simply rendering output of a supercomputer simulation.
Yeah, fine, we'd all like to compute our likelihoods faster, and you can imagine hardware which does it, but it's not clear from the article why doing this analog is superior. Or even how you can avoid sampling algorithms like MCMC using this approach. How does making probabilities analog get you a joint probability distribution over a parameter space, let you marginalize parameters to a lower dimensional distribution, and all the other things MCMC is for? I have a feeling that this is chip is targeted at Bayesian networks, where you're just propagating conditional probability values down a graph, rather than at MCMC.
"Since it is generally impossible to measure what is important, bureaucrats instead turn their energies toward making important what is measurable." — J.M.W. Slack, Egg and Ego
Your position is utterly absurd - it seems to ignore the fact that people can 1) lie and 2) be mistaken.
I didn't say that people can't lie or be mistaken. I explicitly said that people can be mistaken. I also said that is not an excuse to discard eyewitness testimony. It is the judgment of the court whether the testimony provides a reasonable standard of evidence or not. It is your position that is utterly absurd: that no evidence such evidence can ever decide a case.
A single person saying X should not be enough for a court to decide X with NO OTHER INFORMATION WHATSOEVER.
It can be and often is. As I said, criminal trials have been decided on the basis of eyewitness testimony before, and this isn't even a criminal trial. It's up to the judge to decide whether the witness is credible and the testimony likely to be reliable; if so, the testimony can decide the case. If there is some reason to doubt the witnesses credibility -- and there could be, but this is not automatically the case -- then the testimony can be ignored.
It may not be hard to ascertain someone is going 20 mph over the speed limit, but it should not PROVE IT BEYOND A REASONABLE DOUBT.
Yes, it can prove it beyond a reasonable doubt, unless reasonable doubts have been raised about the credibility of the witness. And this isn't a murder trial. As another poster pointed out, minor moving violations don't have to be proven "beyond a reasonable doubt" in the first place, in most courts of law -- they have a lower standard of evidence.
You wouldn't believe conclusively in Aliens if a guy in Iowa said that he saw a UFO out in his field.
Look, you would come across as a lot less stupid if you'd drop these moronic analogies and just argue the case at hand. If you want to argue that there's reason to believe the cop was lying, or unable to adequately judge the speed, then argue that. But a cop visually estimating the physical speed of a vehicle is not even remotely equivalent to a random guy in NYC claiming he saw Jesus.
Jesus.
It's true that there can be unconscious biases in any subjective judgement which renders the judgment uncertain -- that's why it's "subjective". It's always going to come down to details. If it's 20 mph over, then it's fairly certain that you can tell that pretty reliably, regardless of what kind of car they're driving. If it's 5 mph over, then "red-car bias" could play a role, and that's a matter for the judge to decide. I'm just saying that "subjective judgement" isn't the same thing as "no evidence", as the original poster claimed.
Eyewitness testimony certainly can be, and often has been the basis of conviction, even for captial murder cases. If one guy stabs another guy right in front of you, it's not really difficult to ascertain that actually happened and wasn't a hallucination, provided the light is good, you got a good look at the guy or know him, etc.
Likewise, it's not hard to ascertain that someone is, say, going 20 mph over the speed limit just by looking.
Your attempt to conflate "he's was obviously going way over the speed limit" with "it appeared he turned into a flaming dragon" only shows how absurd your position actually is. In the real world, people are able to judge physical events which happen in front of them, such as distances, speeds, or actions.
Certainly this is not perfect -- a person can't tell reliably if someone is 1 mph over the speed limit -- and in the case discussed here, there is room for debate.
But that doesn't change the fact that it is totally ridiculous for you to claim that eyewitness testimony about physical events doesn't constitute evidence or that it's equivalent to belief in the Second Coming. Just give up and stop embarrassing yourself. You're replacing a bad analogy with a ludicrous one.
"Beyond all reasonable doubt" is only required of criminal cases, which this is not, as another poster pointed out. For a moving violation such as speeding, many courts only require a lower standard, such as "clear and convincing evidence". So, again, the question is whether subjective evidence meets this standard, in this case.
It's certainly possible to eyeball a car and, say, tell it's going 20 mph over the limit. In this particular case, the details are more ambiguous. My point is not that eyewitness testimony rates an automatic conviction. It's merely that eyewitness testimony can be sufficient for conviction, depending on circumstances, and that relying on eyewitness testimony is not equivalent to convicting somebody because he looks guilty, ignoring facts and evidence, or any of the ridiculous hyperbole of the original poster.
Of course subjective opinion is not as good as objective measurement. That's not the question. The question is whether subjective evidence, in this case, is sufficient to reasonably ascertain guilt.
To return to the previous poster's murder analogy theme: photographic evidence of one person stabbing another is certainly better than eyewitness testimony. Eyewitnesses can be mistaken. However, in the absence of someone happening to catch the murder on camera, eyewitness testimony itself may be sufficient to convict. It depends on the credibility of the eyewitness as unbiased and an assessment of whether the judgment of the witness is likely to be reliable (e.g., what were the lighting conditions, did the eyewitness have a good line of sight, etc.) That's what needs to be discussed here, not whether subjective judgement is equal to a radar gun.
That's not the spirit of it at all. The ruling is that eyewitness testimony can constitute reasonable evidence, albeit evidence which is not as strong as, say, photographic proof. It's bizarre that you can't understand this. Are you also railing against eyewitness testimony in murder trials? The real issue here is not that "appearances trump facts". The real question is whether eyewitness testimony about vehicle speeds constitutes sufficient evidence. This can be debated reasonably without ridiculous claims that "appearances are more important than facts".
It's not analogous to condemning a person for "looking wrong". It's eyewitness testimony as evidence of a person's actions: "It looked like you were speeding" is analogous to "It looked like you stabbed that guy". Yes, eyewitness judgment can be wrong, but eyewitness judgment is not the same as "you look evil therefore you are guilty".
"You look like a murderer" is more analogous to "you look like a speeder". It is quite different from "it looked like you were speeding", and has nothing to do with the case being discussed here.
Be careful with DIY ECG/EKG. You don't want to mess up and accidentally run too much current through your heart. Be sure that you use an optical decoupler to isolate the power source from the detector. (The way this works, IIRC, is you turn the electrical signal into light using an LED, then use a photodetector to convert that back into electricity, so there is no direct path of conduction between your heart and the ECG.)
Compressed sensing is the same mathematics behind the Rice single pixel camera covered on Slashdot a few years ago.