Discuss
Topics
I thought this might help you Alex ;)
Love it!
I'm pleased to see some reasonable responses to that web site...
I think your whole blog is sort of dishonest. You pick some objection, refine it to the level of absurd, and then "successfully" rebutt your own absurd formulation. I don't believe that many reasonable people would object to use of computer-aided experiments, especially the one that crunch millions of state variables around a realistic globe. The problem is that it looks like many modelers have wrong understanding of what kind of system they are dealing with. For example, they have very loose metrics for model validation. I've seen a bunch of isotherms and hot spots they claim to be in good agreement with historical data. I would ask, by what measure?
Actually, I have a contribution to make, maybe a couple of PhDs can be churned out of this. You are asking, "how can we hope to test a 100 year temperature projection?". As I understand, the current method is to take some scarce data from, say, 1890, and run calculations forward for 100 years, and compare the result with complete well-known (sort of) 1990 data.
Here is the idea: (a) the GCMs are dynamical systems built mostly from first principles, Navier-Stokes equations, radiation, diffusion, and such. (b) it seems that they run into sensitivity to initial condition issue; (c) dynamical systems are reversable in time. Therefore, instead of starting from inaccurate initial data from 1890 or whatever, one need to start from today's accurate data, and calculate the model in REVERSE (provided that computer algorithms are properly implemented). Given much more accurate initial data, it is more likely that climate conditions will be restored more accurately back in time. If not, then this approach should be able to establish limits for climate forecast time.I think your whole blog is sort of dishonest. You pick some objection, refine it to the level of absurd, and then "successfully" rebutt your own absurd formulation. I don't believe that many reasonable people would object to use of computer-aided experiments, especially the one that crunch millions of state variables around a realistic globe. The problem is that it looks like many modelers have wrong understanding of what kind of system they are dealing with. For example, they have very loose metrics for model validation. I've seen a bunch of isotherms and hot spots they claim to be in good agreement with historical data. I would ask, by what measure?
Actually, I have a contribution to make, maybe a couple of PhDs can be churned out of this. You are asking, "how can we hope to test a 100 year temperature projection?". As I understand, the current method is to take some scarce data from, say, 1890, and run calculations forward for 100 years, and compare the result with complete well-known (sort of) 1990 data.
Here is the idea: (a) the GCMs are dynamical systems built mostly from first principles, Navier-Stokes equations, radiation, diffusion, and such. (b) it seems that they run into sensitivity to initial condition issue; (c) dynamical systems are reversable in time. Therefore, instead of starting from inaccurate initial data from 1890 or whatever, one need to start from today's accurate data, and calculate the model in REVERSE (provided that computer algorithms are properly implemented). Given much more accurate initial data, it is more likely that climate conditions will be restored more accurately back in time. If not, then this approach should be able to establish limits for climate forecast time.
Hmmm...at first I thought "yeah! That is genius! Run the model in reverse!" But, running these models, and the complex factors of cause and effect that they model, in "reverse", is not like rewinding a tape, and watching everything happen exactly as it would in a progression in time, only in reverse.
If you've ever created computer models of complex systems, dynamics, and interactions, you'd know that flipping the algorithms, reversing the many factors, and "running the model backwards" based off of current data will not produce a snapshot of climate conditions 100 years ago. The peson quoted above writes "provided the computer algorithms can be properly implemented", but creating a real climate model that can model climates in reverse would be far more difficult than creating a climate model that can run climates in the normal direction of time. Climates are highly complex systems, that don't happen in "reverse". What emerged from climate systems 100 years ago was the product of of climate change moving through time in the normal direction. Polar ice that has melted between 100 years ago and now would not automatically re-establish itself in the same place if climate models we designed to run in reverse. That is because this polar ice has been in it's place for tens of thousands of years, and may have taken many tens of thousands of years to build up where it did. So, in a place where ice no longer exists, even if algorithms are made to run climate simulations in reverse, great collections of ice will not automatically re-establish in 100 years in the same place where they melted away in 100 years. Sorry, it complex systems do work like VCR's
Yep, I agree. Models can only give you indications and actually the best thing they are for is trying to understand the independencies. But if you need to understand the interaction of the parts, then you can't really create a great model. You can't make some assumptions to test the assumptions.
Can I say it? Garbage in, garbage out.
I would also say that most of the real evidence for the case for global warming comes not from climate models, but from Paleoclimatelology. Computer climate models are supplementary tools. They are not the foundation of climate science.
That's a good point Sam. Though I'd wish climate models were not used no widely in arguments regarding climate decisions/exposure.
You 're right. It would actually be great to see some kind of online, neutral-territory, open debate forum, where people from all sides of the argument can lay out their evidence in a clear, transparent, and open way. This issue is so muddled, spun, politicized, and clouded with loaded arguments by agenda-backed think tanks from both sides. It's become one of those issues that has people so polarized that they are walkking around in two different realities: one reality where we are experiencing human influenced global warming climate change, and one where we are not. Yet, somewhere among all of the "sides" in the argument is the truth.
Basically, IMO, if human influenced global warming is real, then we need to do something about it. And, if it isn't real, then we need to examine who is behind creating the myths about it and why. It would be cool to see issues like this argued in some kind of neutral territory, and the transcripts of the argument made public. sort of like a court hearing where evidence is presented from both (or all) sides. At least, more preferrable IMO than seeing the arguments played out in the media, and through possibly biased academic institutions, and possibly paid off ad questionable think tanks. An issue lik ethis is far too important to be debated in the types of forums that it is debated within now.
I think it will require a news headline something like: Breaking News Via Google - Weather is Gone! for a lot of people to admit that there is any problem more complex than their cable bill is due. I am sad to report that about 30 percent of the people online equate global warming to a statement by their teacher that it is summer-time! The problem is made much worse by "scientists" with hidden agendas. I used to work with one who was honest about 12 years ago, and we studied all these eventualities back then. Pretty soon we are going to have to break out the plunger on mother Earth, as we will have a real problem flushing these problems away.
