Global Warming Models

[ponderingturtle]

Robustness comes from peer reviewed examinations such as I linked to. There are additional papers available.

No nessacarily. You can review a model all you want, and it will not be robustly tested untill it is applied in different situations and shown to make useful predictions. You can have a model that works for the last 100 years perfectly, but that does not mean that the model is particularly valid. just that some aproximations and such which might not generaly hold, hold for the period being discussed.

So over limited data sets do restrict how robust any model can actualy be tested.

The Real Climate blog is not intended be a completely through research tool. That link was to provide an overview of how they validate. The second link was to provide an example where they validated models against longer time spans than the real climate information talked about.

The models require initial conditions on a great number of variables to begin modeling, more than just temperature. We don't have those numbers so it's hard to initialize the model. It's possible to come up with a set of numbers that make the model work but with nothing to compare against you can't say for sure that the initial conditions were valid.

It has just struck me that the data was highly limited and that would restrict how robust the model can be made. And that was what I was wondering how it was being looked at.

 

[ponderingturtle]

All scientific models and theories are based on observation, in this case of the past climate. A model is then produced based on this to try to predict the future conditions. If a global warming model failed to predict the little ice age, it would be disagreeing with the very data it was based on and would be completely worthless. The only way to test a scientific model is to first ensure it agrees with the known data, in this case the past, and then test it's predictions for the future.

Not really. You can test it with historic data, it is just data not orrigionaly used in building the model. If you say used the past 150 years as the data set for you model, a way to test it would be look at say the 450 years before your data set.

You do not need to look to the future to test a model, looking at historic hapenings is just as good. My understanding is that alot of say geology is based on this method

 

[ponderingturtle]

http://www.abc.net.au/rn/scienceshow/stories/2006/1685580.htm#

If the models don't convince you, then the events in progress might.

Melting glaciers and permafrost, for example.

The models are just trying to say what will probably happen, but the warming is clearly evident without them. Unless some magic negative feedback mechanism appears, warming is what it will be.

Well I have heard of the deep ocean coveyor shutting down form sudden mess up in salinity being a strong negative feedback mechanism. Also I don't hear much about how we are thought to be in an ice age, just in an interglacial period.

 

[Dave1001]

Not really. You can test it with historic data, it is just data not orrigionaly used in building the model. If you say used the past 150 years as the data set for you model, a way to test it would be look at say the 450 years before your data set.

You do not need to look to the future to test a model, looking at historic hapenings is just as good. My understanding is that alot of say geology is based on this method

I see Cuddles point though. You're just describing being partially through with designing the model. Becaude if the model doesn't work with 450 year old date, one tweaks the model so that it works with both 150 year and 450 year old data, right? And so on and so on until there is no data left to work with. So now you have your finished model. Still won't know if it works until you see how well it predicts the future.

 

[ponderingturtle]

I see Cuddles point though. You're just describing being partially through with designing the model. Becaude if the model doesn't work with 450 year old date, one tweaks the model so that it works with both 150 year and 450 year old data, right? And so on and so on until there is no data left to work with. So now you have your finished model. Still won't know if it works until you see how well it predicts the future.

So we what do we need to see happen to prove other historic ideas? By this idea evolution of large animals would never have been tested in any real fashion becuase the time scale has not been right. But it makes definite predictions as to the nature of historic events and as more information is found out these can either support the theory/model or not.

You do not need to predict the future to test a model, just increase the data set in any fashion.

What is so different about tweaking a model because recent events did not fit it and tweaking a model because historic events did not fit it? I do not see why one is being valued over the other. Now this case makes it hard because much of the historic data is not of high quality but if it is good enough it should be able to be used in testing

 

[CapelDodger]

Well I have heard of the deep ocean coveyor shutting down form sudden mess up in salinity being a strong negative feedback mechanism.

Not that strong, since the effect will be mostly limited to the North Atlantic. The heat that wouldn't be transported there would still exist, to stagnate or be transported elsewhere, so the global effect would be far less than the regional effect. The global cooling effect would be the increased albedo caused by advancing glaciers and longer-lasting snow-fields around the North Atlantic and in the Arctic Ocean
.

Also I don't hear much about how we are thought to be in an ice age, just in an interglacial period.

We are in an ice-epoch, which is characterised by repeated ice-ages separated by interglacials (one of which is happening now).

 

[CapelDodger]

It has just struck me that the data was highly limited and that would restrict how robust the model can be made. And that was what I was wondering how it was being looked at.

Models have been around for several decades, with good data, and they have predicted the present (their future) remarkably well.

 

[ponderingturtle]

Models have been around for several decades, with good data, and they have predicted the present (their future) remarkably well.

But when does that cross the line into being actualy rigorous or have they been massaging the numbers as they go and changing the models like people where talking about doing with past data here?

 

[CapelDodger]

But when does that cross the line into being actualy rigorous or have they been massaging the numbers as they go and changing the models like people where talking about doing with past data here?

Model predictions from the 80's have stood up well to the test of time. They haven't been altered subsequently, they are what they were. If anything they've rather understated the rate of warming.

Models aren't "massaged", they're improved in the light of experience. When reality diverges from model predictions it reveals errors in either the programming or the underlying physics, errors which can then be addressed. It's not simply a question of "changing the numbers" and hoping nobody notices. If (for instance) the physics of cloud-formation as originally understood and incorporated in a model turn out to be wrong, the process can be further researched and a better understanding arrived at. This better understanding can then lead to a better model. That's the prime function of most models (not just in climate science) : to determine how good current theory is, identify any failings, and improve the physics.

 

[kevin]

No nessacarily. You can review a model all you want, and it will not be robustly tested untill it is applied in different situations and shown to make useful predictions. You can have a model that works for the last 100 years perfectly, but that does not mean that the model is particularly valid. just that some aproximations and such which might not generaly hold, hold for the period being discussed.

Again, what I gave was a limited overview. The predictions are out there. The models predicted increased snow fall on interior ice sheets would cause those sheets to thicken. This has been shown to be happening, another success for the models.

BTW the models are not date/time predictors. They tend to be CO2 level predictors. So they predict when the CO2 reaches this level, this will occur. The date/time predictions come from forecasting out current CO2 emissions. If something changes that emission levels, positively or negatively, the time line changes but the results the models predict have still been accurate.

It has just struck me that the data was highly limited and that would restrict how robust the model can be made. And that was what I was wondering how it was being looked at.

They do this by generating addtional datasets. This is an ongoing process. Currently there are data sets for tree ring data, ice core data, corel reef data, observation data, satellite data. Models are re-evaluated to new datasets as they appear.

 

[kevin]

Models aren't "massaged", they're improved in the light of experience. When reality diverges from model predictions it reveals errors in either the programming or the underlying physics, errors which can then be addressed.

Or errors in the datasets! This has happened too. Most recently with satellite data.
http://www.realclimate.org/index.php/archives/2005/08/et-tu-lt
http://www.realclimate.org/index.php/archives/2005/11/more-satellite-stuff

 

[ponderingturtle]

Again, what I gave was a limited overview. The predictions are out there. The models predicted increased snow fall on interior ice sheets would cause those sheets to thicken. This has been shown to be happening, another success for the models.

BTW the models are not date/time predictors. They tend to be CO2 level predictors. So they predict when the CO2 reaches this level, this will occur. The date/time predictions come from forecasting out current CO2 emissions. If something changes that emission levels, positively or negatively, the time line changes but the results the models predict have still been accurate.



They do this by generating addtional datasets. This is an ongoing process. Currently there are data sets for tree ring data, ice core data, corel reef data, observation data, satellite data. Models are re-evaluated to new datasets as they appear.

The thing is if the model is being revised through new data and corrected in that way, why should its successful predictions be counted as accurate? The unsuccessful ones would be atributed to bad data or not enough, but the successful one would not be.

It just feels a bit like a practitioner of Woo choosing to show the patients that got better from his BS and saying that it then counts as science.

That is the feel I have gotten from alot of enviromental related statements. I wonder what data is used to gain an average extiction rate to compare the current one to and so on. So I get a feeling of a selectivity bias being introduced in many of these kinds of predictions

 

[Cuddles]

So we what do we need to see happen to prove other historic ideas? By this idea evolution of large animals would never have been tested in any real fashion becuase the time scale has not been right. But it makes definite predictions as to the nature of historic events and as more information is found out these can either support the theory/model or not.

You do not need to predict the future to test a model, just increase the data set in any fashion.

What is so different about tweaking a model because recent events did not fit it and tweaking a model because historic events did not fit it? I do not see why one is being valued over the other. Now this case makes it hard because much of the historic data is not of high quality but if it is good enough it should be able to be used in testing

Evololution hasn't been thouroughly tested. Looking at fossils led to a theory of evolution by natural selection. Looking at more fossils only supports the theory or leads to improvements.

It is not true that there has not been enough time to teat predictions, since tests on fast breeding animals such as mice can procede on a human timescale. In fact, all research into drug resistant bacteria is effectively testing the predictions of evolutionary theory.

 

[ponderingturtle]

Evololution hasn't been thouroughly tested. Looking at fossils led to a theory of evolution by natural selection. Looking at more fossils only supports the theory or leads to improvements.

That depends on what you define as the being part of it vs not. The difference here is that while many of the fundamental parts of it are not at issue(I don't think anyone questions the green house effect) it is the detailed outcomes that are being questioned. With evolution it would be more like stateing that CO2 does not increase heat retained by the atmosphere. That would be a similar discrepency to get it to the level of the popular debate on evolution

 

[Cuddles]

The thing is if the model is being revised through new data and corrected in that way, why should its successful predictions be counted as accurate? The unsuccessful ones would be atributed to bad data or not enough, but the successful one would not be.

It just feels a bit like a practitioner of Woo choosing to show the patients that got better from his BS and saying that it then counts as science.

That is the feel I have gotten from alot of enviromental related statements. I wonder what data is used to gain an average extiction rate to compare the current one to and so on. So I get a feeling of a selectivity bias being introduced in many of these kinds of predictions

I have to admit I am also sceptical about many environmental statements like this. Not so much because of selectivity, but because of the lack of accurate data. We know how few animals become fossilised, yet people make statements about extinction rates in the past, even though we don't know how many species there were at the time. Or now for that matter.

This is why predictions of the future are more important than past data. Past data allows you to adapt your model without showing up errors. If you make a prediction that turns out to be wrong you can have no excuses other than an inaccurate model or data.

 

[drkitten]

The thing is if the model is being revised through new data and corrected in that way, why should its successful predictions be counted as accurate? The unsuccessful ones would be atributed to bad data or not enough, but the successful one would not be.

Depends on the ratio of successful to unsuccessful predictions, doesn't it?

Suppose that I've been following (and predicting) the American baseball season so far, and I've been able to predict with about 90% accuracy who will win any given game. Am I "accurate" or not? I'm sure that I'm far too accurate for my bookie's comfort --but there's still room for improvement.

The standard (Guthrie heel prick) test for pheylketonuria has a false positive rate of about 0.2%; its false negative rate is about 0.3%. Is this test "accurate"? Yes, but there's still research going on to make it better.

 

[Cuddles]

That depends on what you define as the being part of it vs not. The difference here is that while many of the fundamental parts of it are not at issue(I don't think anyone questions the green house effect) it is the detailed outcomes that are being questioned. With evolution it would be more like stateing that CO2 does not increase heat retained by the atmosphere. That would be a similar discrepency to get it to the level of the popular debate on evolution

And yet this hasn't been proved. It is very likely that CO2 increases warming, and the past data suggests this. The only way to actually test the models is to predict what the temperature will be in the future given a certain level of CO2 and then wait around to see what happens. I think the difference between evolution and global warming is that evolution conflicts with some people's personal belief that they are unwilling to give up, whereas global warming is a purely scientific argument. The general public may believe or not in global warming, but all their opinions stem from what the media tells them and not from their own beliefs.

 

[ponderingturtle]

Depends on the ratio of successful to unsuccessful predictions, doesn't it?

Suppose that I've been following (and predicting) the American baseball season so far, and I've been able to predict with about 90% accuracy who will win any given game. Am I "accurate" or not? I'm sure that I'm far too accurate for my bookie's comfort --but there's still room for improvement.

The standard (Guthrie heel prick) test for pheylketonuria has a false positive rate of about 0.2%; its false negative rate is about 0.3%. Is this test "accurate"? Yes, but there's still research going on to make it better.

So why do we never hear confidence numbers on global warming predictions? That would make me more apt to trust the seemingly random numbers spouted by various groups.

 

[ponderingturtle]

And yet this hasn't been proved. It is very likely that CO2 increases warming, and the past data suggests this.

I was specificaly refering to the greenhouse effect, not how that effect will change the climate. I got the impression that there was a difference between the two.

The only way to actually test the models is to predict what the temperature will be in the future given a certain level of CO2 and then wait around to see what happens. I think the difference between evolution and global warming is that evolution conflicts with some people's personal belief that they are unwilling to give up, whereas global warming is a purely scientific argument. The general public may believe or not in global warming, but all their opinions stem from what the media tells them and not from their own beliefs.

The other problem is that alot of enviromentalism are not making arguements from a scientific perspecitive. They both have lots of woo involved it is just classical woo vs newage woo.

Evolution has been around and tested for longer as well, and that is a seperate issue. And I do see that while details of one are not quite rated the same between them

Example

Evolution
Basic Premise-Species change into new species
Detail-Cats and Dogs sperated as groups 50 million years ago

Climate Change
Basic Premise-The climate can undergo drastic changes
Detail- In 100 years the sealevel will be 10 feet higher than now

The basic premise of evolution is what people have problems with, and while the details change it does not change the basic premise

The basic premise of climate change poeple do not generaly have issues with, but the detailed predictions and such are where the discrepencies occur.

So compairing the two and the issues they have in the popular media is not accurate. I was questioning a detailed part of the theory regarding predicting climate change. It would be like say questioning when the last common ancestor that humans and chimps have is 3 million years ago or 7 million years ago. If one could be shown it would change the details of evolutionary theory but not the basic premise, but with climate change it is the details we are interested in.

 

[varwoche]

So why do we never hear confidence numbers on global warming predictions? That would make me more apt to trust the seemingly random numbers spouted by various groups.

While I have no doubt there are endless agenda-driven web sites containing all sorts of gibberish, I'm wondering if you can point to examples of "seemingly random numbers" published by known, seemingly credible sources?