Scepticism Of Man Made Climate Change

[VillagePlank]

 In this case (climate models) what counts is the 'scientific correctness'.

 

... which is the mathematical models mapped to a machine which requires machine correctness!

[VillagePlank]

Generally climate models don't run off initial conditions in the way weather models do. Some even start with the atmosphere in a very simple state (isotherm, at rest, no vegetation, etc) and require a spin up time before they resemble a realistic atmosphere.There is a range of reasons why one would want to run an ensemble, because there is a range of things that can be varied and still produce a plausible result, e.g. the correct settings for some parametrisation, the grid resolution, different types of parametrisations, time steps, etc.But even with settings you do vary, you are not concerned about a difference in the 10 decimal place, not even in weather modelling.

 

Well, I would be concerned; particularly given Lorenz's experience with differences at the 6th decimal place. Re chaos: see Hamill and Colucci (1997,1998) Eckel and Walters (1998) Bremnes (2004), Doblas-Reyes (2005) and Raftery (2005) For extra see Weigel (2009) You might also be interested in reading the findings of Sanders (1973), Bosart (1975) and Gyakum (1986). You might also be interested in doing an experiment with the logistic map changing decimal points at the 12th and 13th decimal place and seeing what happens.

Maybe not to write the code to control a 747, though

 

Exactly; you'd satisfy risk when it's your bottom on a seat on a plane, but not when the future climate and economic well being of the entire human population is concerned. Sums it up nicely. Thanks.

 

It remains my view that scientists shouldn't be writing code.

Edited August 19, 2013 by Sparkicle

[VillagePlank]

 In my, certainly limited, experience it's the CompSci people, the professional programmers without any domain knowledge that often deliver too late and over budget

 

Well, according to the IEEE, these are the primary reasons software projects fail:

 

[*]Unrealistic or unarticulated project goals

[*]Inaccurate estimates of needed resources

[*]Badly defined system requirements

[*]Poor reporting of the project's status

[*]Unmanaged risks

[*]Poor communication among customers, developers, and users

[*]Use of immature technology

[*]Inability to handle the project's complexity

[*]Sloppy development practices

[*]Poor project management

[*]Stakeholder politics

[*]Commercial pressures

http://spectrum.ieee.org/computing/software/why-software-fails

[VillagePlank]

That is not to say that you can ignore the pitfalls of computational mathematics and floating point representations, but modellers are generally aware when to be worried about those issues.

 

Well, show me a climate scientist that has read this. The end result is, if you want your software to be (mathematically) correct (your definition), you have to go through immense amount of pain, anguish, and gnashing of teeth, or, instead, use rational numbers. and in the worse case scenario use interval arithmetic so that you know your computational error bounds.

 

This has now gone well off-topic, so I'll leave it at that.

Edited August 19, 2013 by Sparkicle

[VillagePlank]

The IPCC suggest that there is a 'non-linear, chaotic nature of the climate system' So care about correctness at the nth decimal place due to initial sensitivty seems to be well warranted. Well, at least, in my world it is.

[SomeLikeItHot]

Sparkicle, on 19 Aug 2013 - 08:03, said:Well, I would be concerned; particularly given Lorenz's experience with differences at the 6th decimal place. Re chaos: see Hamill and Colucci (1997,1998) Eckel and Walters (1998) Bremnes (2004), Doblas-Reyes (2005) and Raftery (2005) For extra see Weigel (2009) You might also be interested in reading the findings of Sanders (1973), Bosart (1975) and Gyakum (1986). You might also be interested in doing an experiment with the logistic map changing decimal points at the 12th and 13th decimal place and seeing what happens.

The issue is signifigant figures or the percentage error that any rounding may create not decimal places per se. The accuracy you are recording data to at a particular location in a model may not need to be high. For example if your grid points are 100 miles and 3 hours apart on your model, recording a temperature at 12.46 or 12.5 degrees at 6am is irrelevant we know the weather models don't get that kind of accuracy one day ahead yet they continue to produce reasonably accurate predictions 4-5 days ahead. Even with non-linearities at some level more precision is essentially spurious precision. You don't need to calculate at maximum accuracy but just ensure that any errors that this truncation will create will grow more slowly than errors introduced by your other modelling assumptions.

QuoteExactly; you'd satisfy risk when it's your bottom on a seat on a plane, but not when the future climate and economic well being of the entire human population is concerned. Sums it up nicely. Thanks.It remains my view that scientists shouldn't be writing code.

Or programers working out correct modelling assumptions?

[VillagePlank]

The issue is signifigant figures or the percentage error that any rounding may create not decimal places per se. The accuracy you are recording data to at a particular location in a model may not need to be high. For example if your grid points are 100 miles and 3 hours apart on your model, recording a temperature at 12.46 or 12.5 degrees at 6am is irrelevant we know the weather models don't get that kind of accuracy one day ahead yet they continue to produce reasonably accurate predictions 4-5 days ahead. Even with non-linearities at some level more precision is essentially spurious precision. You don't need to calculate at maximum accuracy but just ensure that any errors that this truncation will create will grow more slowly than errors introduced by your other modelling assumptions.Or programers working out correct modelling assumptions?

NO.

 

See Knuth for floating point inaccuracies along with proofs. Best bet for you is to try again. I can't be bothered to go through some of the (difficult) maths with you. Sorry. I note the 'likes' you get. Amusing to say the least.

Edited August 19, 2013 by Sparkicle

[BornFromTheVoid]

 I note the 'likes' you get. Amusing to say the least.

 

...errr, whatever floats your boat!

 

 

Exactly; you'd satisfy risk when it's your bottom on a seat on a plane, but not when the future climate and economic well being of the entire human population is concerned. Sums it up nicely. Thanks.

 

It remains my view that scientists shouldn't be writing code.

 

Without the computer models, warming was still predicted. So it's only part of the research that leads to policy suggestions and their potential economic influences.

 

How's about some links to the papers you've mentioned?

[VillagePlank]

Oh do give it a rest, BFTV. Pay for the info like I have to. How about a comment on using linear regression to crush your enemies?

[VillagePlank]

Without the computer models, warming was still predicted.

 

Was it? Are you sure?

[VillagePlank]

Annoyed I am; it appears no one reads what I say. I am going on a self induced holiday. Cheers, have fun, and stay safe xxx

 

EDIT: I apologise if I have offended anyone. Really, and honestly, I am only conducting honest and open inquiry. It is clear that I am often wrong, and flag it up myself as soon as I can. I know a lot of people do not like such things.

Edited August 19, 2013 by Sparkicle

[knocker]

Was it? Are you sure?

 

On increasing global temperatures: 75 years after Callendar

 

In 1938, Guy Stewart Callendar was the first to demonstrate that the Earth’s

land surface was warming. Callendar also suggested that the production of carbon

dioxide by the combustion of fossil fuels was responsible for much of

this modern change in climate. This short note marks the 75th

anniversary of Callendar’s landmark study and demonstrates that his global land temperature

estimates agree remarkably well with more recent analyses

 

http://www.met.reading.ac.uk/~ed/home/hawkins_jones_2013_Callendar.pdf

Edited August 19, 2013 by knocker

[john pike]

http://www.nipccreport.org/articles/2013/aug/7aug2013a1.html

 

http://www.nipccreport.org/articles/2013/aug/14aug2013a2.html

Edited August 19, 2013 by john pike

[loafer]

On increasing global temperatures: 75 years after Callendar

 

In 1938, Guy Stewart Callendar was the first to demonstrate that the Earth’s

land surface was warming. Callendar also suggested that the production of carbon

dioxide by the combustion of fossil fuels was responsible for much of

this modern change in climate. This short note marks the 75th

anniversary of Callendar’s landmark study and demonstrates that his global land temperature

estimates agree remarkably well with more recent analyses

 

http://www.met.reading.ac.uk/~ed/home/hawkins_jones_2013_Callendar.pdf

That isn't particularly surprising, given the models will have been played with until their outputs match historic actuals closely.

However, the problem with predictions is that they involve looking into the future, which is a bit harder, as we have seen and, as with stocks and shares, past performance is no indication of future performance.

[UV-RAY]

 

On increasing global temperatures: 75 years after Callendar

 

In 1938, Guy Stewart Callendar was the first to demonstrate that the Earth’s

land surface was warming. Callendar also suggested that the production of carbon

dioxide by the combustion of fossil fuels was responsible for much of

this modern change in climate. This short note marks the 75th

anniversary of Callendar’s landmark study and demonstrates that his global land temperature

estimates agree remarkably well with more recent analyses

 

http://www.met.reading.ac.uk/~ed/home/hawkins_jones_2013_Callendar.pdf

 

Indeed they were rising, but look at the date of Callanders study and the data for back then.

[knocker]

However, the problem with predictions is that they involve looking into the future, which is a bit harder, as we have seen and, as with stocks and shares, past performance is no indication of future performance.

 

But I've read a number of peer reviewed papers that state the more we understand past climate changes may well give us a greater understanding of the current situation. But still as the great Niels Bohr said, "Prediction is very difficult, especially if it's about the future".

[loafer]

But I've read a number of peer reviewed papers that state the more we understand past climate changes may well give us a greater understanding of the current situation. But still as the great Niels Bohr said, "Prediction is very difficult, especially if it's about the future".

Your statement is completely compatible with mine.I am simply saying that creating a model which can hindcast with a high correlation with actual events does not mean you have created a model which can forecast accurately, which is what the article was trying to say.

[SomeLikeItHot]

NO. See Knuth for floating point inaccuracies along with proofs. Best bet for you is to try again. I can't be bothered to go through some of the (difficult) maths with you. Sorry. I note the 'likes' you get. Amusing to say the least.

You clearly know and care a lot about programming issues, no one is denying that issues such as floating point accuracy can throw out calculations, this is probably one of the more wll known issues in scientific computing. Nor that bad coding standards will lead to difficult to maintain and test code and more frequent errors.The question is not to get hung up on some technical point just because its the one you understand, but rather look at if it is the relevant error here? We have measurement error in our inputs, we have sampling error in our inputs, we have modelling error from the necessarily coarse granularity of the calculations. All these will rapidly dwarf any error in the 10th decimal place from a floating point error.

[keithlucky]

UN IPPC"s Kevin Trenbeth oceans will eat global warming for another 20yrs .quote

The oceans can at times soak up a lot of heat. Some goes into the deep oceans where it can stay for centuries. But heat absorbed closer to the surface can easily flow back into the air. That happened in 1998, which made it one of the hottest years on record.

Trenberth says since then, the ocean has mostly been back in one of its soaking-up modes.

“They probably can’t go on much for much longer than maybe 20 years, and what happens at the end of these hiatus periods, is suddenly there’s a big jump [in temperature] up to a whole new level and you never go back to that previous level again,†he says.

[loafer]

See how 150 years of sea level rise will affect US coastal cities in 30 seconds

 

http://www.climatecentral.org/news/150-years-in-30-seconds-sea-level-debt-sinking-us-cities-16338

Not quite...if you look at the notes below, this is a graphic showing when, over the next 150 years, they consider that enough CO2 will be "locked in" to the climate system to result in flooding in those cities at some undetermined point in the further future.

Basically, it is a load of alarmist nonsense.

[keithlucky]

Watch video on A Gore and warmists as mother nature is not listening global warming hysteria http://www.sunnewsnetwork.ca/video/featured/prime-time/867432237001/al-gore-climate-hysteria/2623328888001

 

 

 

[keithlucky]

Earth has gained 2.2million km2 of sea ice this year beating the previous record gain of 1.million km2in 1996.

http://arctic.atmos.uiuc.edu/cryosphere/timeseries.global.anom.1979-2008

[jonboy]

 

Earth has gained 2.2million km2 of sea ice this year beating the previous record gain of 1.million km2in 1996.

http://arctic.atmos.uiuc.edu/cryosphere/timeseries.global.anom.1979-2008

 

Although welcomed it means diddly squat really especially as we were probably 2 million below average. What we what to see is ice loss approaching average on a longer term basis or even reducing. If we have extensive loss and large gains this doesn't improve the state of global ice. I also don't want to significant gains year on year as this would be an indicator of significant cooling which is just as bad as significant warming. I'm just an average type of guy!!!

[keithlucky]

UN climate panel can"t explain warming slowdown http://t.co/k7kA4D42d6

[Methuselah]

UN climate panel can"t explain warming slowdown http://t.co/k7kA4D42d6

Neither can anyone else...